Post-qualification continuing professional development
Observation, analysis and evaluation for public-facing professionals
A facilitator handbook and full curriculum for social work, health,
allied health and related front-line professions
This is written for the person delivering the course as much as for the person taking it. Everything a facilitator needs sits in Part Three. Everything a learner needs to understand why the course is built this way sits in Parts One and Four.
The programme rests on a claim that ought to be stated plainly at the outset. The skills that make Sherlock Holmes an arresting character are not detective skills. They are assessment skills, and they are the same skills that a district nurse uses on a first home visit, that a social worker uses when a chronology will not add up, and that a GP uses in the ninety seconds before a consultation settles into a diagnosis. Doyle happened to dress them in a deerstalker. Strip the costume off and what remains is a working account of how an expert attends, infers, records and decides.
Three warnings before you teach a word of it.
Holmes is a fiction, and fiction is written backwards. Doyle knew the answer before he wrote the clue, which is a luxury no practitioner has ever had. Module 12 takes this apart in detail, and the counterpoint boxes throughout the earlier modules keep seeding the doubt so that the final module is a harvest rather than a surprise.
Holmes works alone, dislikes emotion and reasons about people rather than with them. Every one of those habits is a liability in a caring profession. Where the canon is wrong, this course says so, and says so early.
Inference from appearance is one letter away from prejudice. A course that trains people to read class, race, occupation and circumstance from a coat sleeve is a course that can make discriminatory practice more articulate rather than less. Module 12 addresses this head on, but the facilitator must hold it in mind from Module 1. If a group starts enjoying its own cleverness, that is the moment to intervene.
Canon anchor. The passage from Doyle that the session is built on. Read it aloud. It takes under a minute and it changes the room.
Connection. A link to a field outside health and social care. These are not decoration. Learners retain a principle better when they have met it twice in unrelated clothing.
Exercise. Fully specified, with timings, materials and debrief questions. Numbers assume a group of twelve to eighteen working in threes.
Counterpoint. The objection to the module's own teaching. Deliver it. A course of unbroken admiration produces overconfident practitioners, which is precisely the harm the course exists to reduce.
Workplace task. Done between sessions, evidenced in the portfolio.
The obvious answer is that the method was never fictional to begin with. Conan Doyle was a medical student at Edinburgh in 1877 and served as clerk to Joseph Bell, a surgeon whose party trick was to tell a stranger his occupation, his regiment and where he had lately walked. Doyle wrote to Bell in 1892 that it was most certainly to him that he owed Sherlock Holmes. The character is a doctor's diagnostic manner transposed into crime. Teaching it back to clinicians and social workers returns it to where it came from.
The less obvious answer is that the professions have a measurable problem which this material speaks to. Combined estimates from three large observational studies put outpatient diagnostic error in United States adults at around five per cent, roughly one adult in twenty each year, with about half of those errors carrying potential for harm. In English general practice, record review has found missed opportunities in diagnosis at a rate that is not reassuring either. Serious case reviews and safeguarding adults reviews have been telling the same story from the other direction for thirty years, and they have converged on a phrase which now appears in almost every local training plan. Professional curiosity.
The phrase is the problem. Professional curiosity is usually taught as a disposition, something a practitioner either brings to work or does not, and the training that follows is largely exhortation. Be curious. Ask one more question. Think the unthinkable. None of this tells anyone how. Holmes is useful precisely because he is specific. He does not tell Watson to be curious. He tells him that there are seventeen steps, that the mud on a boot is a particular red clay, that a hat has been brushed on one side only. The canon offers a set of nameable, drillable operations, and nameable operations can be practised, assessed and supervised. That is the whole argument for this curriculum.
There is a further reason, and it is worth stating because it will otherwise be dismissed as charm. The stories are enjoyable. Post-qualification CPD in the public sector competes with a caseload, and it usually loses. Material that people want to read carries training weight that a slide deck on assessment frameworks does not. Use that. Do not apologise for it.
Twenty-one competencies drawn from the canon, grouped into five families, each stated as something a practitioner does rather than something a practitioner has.
The grouping matters more than the count. Perception without inference produces the practitioner who notices everything and concludes nothing. Inference without organisation produces the practitioner who is brilliant in the moment and useless in a case conference eighteen months later. The families are sequenced in the order in which a case actually moves, and the modules follow that order.
"You have frequently seen the steps which lead up from the hall to this room." "Frequently." "How often?" "Well, some hundreds of times." "Then how many are there?" "How many? I don't know." "Quite so! You have not observed. And yet you have seen. That is just my point. Now, I know that there are seventeen steps, because I have both seen and observed." A Scandal in Bohemia
Holmes draws a distinction that psychology arrived at independently sixty years later. Seeing is passive and cheap. Observing is an act of deliberate allocation, and attention is a finite resource that has to be spent on purpose. What Holmes never says, but demonstrates in every story, is that he decides in advance what he is going to attend to. He has a sweep. He does the hands, the cuffs, the boots, the knees, the pockets, in an order, every time.
For practice this converts into something unglamorous and highly teachable. An observation protocol. The professional who enters a home with a fixed sweep sees more than the professional who enters intending to be observant.
Holmes can identify the anomalous only because he has an unusually detailed model of the typical. He knows what a clerk's cuff normally looks like, so a worn one speaks. Anomaly detection is parasitic on a norm, and the norm has to be learned before the anomaly can be perceived at all.
This reframes professional curiosity completely. Curiosity is not an attitude, it is a knowledge base. A health visitor who has seen four hundred kitchens will notice the fifth hundred one is wrong. A newly qualified worker who has seen twelve will not, however curious she has been told to be. The training implication is that we should be building normative libraries and not delivering attitude workshops.
"Is there any point to which you would wish to draw my attention?" "To the curious incident of the dog in the night-time." "The dog did nothing in the night-time." "That was the curious incident." Silver Blaze
The single most transferable idea in the canon. Absence is data, and absence is invisible unless you have gone looking for it. The appointment not attended, the sibling never seen, the medication not collected, the relative who is always out, the bruise nobody has documented. Serious case reviews return to this pattern with dismal regularity, and it is precisely the class of evidence that ordinary attention cannot supply, because human perception is built to register events rather than non-events.
"By a man's finger-nails, by his coat-sleeve, by his boot, by his trouser-knees, by the callosities of his forefinger and thumb, by his expression, by his shirt-cuffs, by each of these things a man's calling is plainly revealed." A Study in Scarlet, "The Book of Life"
Objects and environments carry the history of their use. Holmes reads a watch in The Sign of Four, a hat in The Blue Carbuncle, a stick in The Hound of the Baskervilles. Each time the move is identical. He asks what pattern of use would have produced this wear.
The clinical and social care equivalents are everywhere and are usually under-taught. The state of a fridge. Which chair in a living room has been sat in. Whether the child's bed has been slept in. Whether the dosette box has been filled by someone who understands it. Wear patterns on a walking frame. This is the most enjoyable family to teach and the most dangerous, for reasons Module 12 sets out.
Holmes calls it deduction. It is not. Deduction guarantees its conclusion from its premises, and Holmes never has premises of that kind. What he actually performs is what Charles Sanders Peirce called abduction, inference to the best available explanation, which is defeasible by nature. A tan and a wounded arm are consistent with Afghanistan and also with a dozen other histories.
Getting this right is not pedantry, it is the epistemic spine of the whole programme. A conclusion reached by abduction must be held provisionally, must be stated with its competitors visible, and must be revisable on new information. A conclusion reached by deduction need not. Practitioners who believe they are deducing will not revise. This is the mechanism by which a working hypothesis hardens into a case narrative that nobody can shift, and it is one of the recurring findings of safeguarding reviews.
"Most people, if you describe a train of events to them, will tell you what the result would be... There are few people, however, who, if you told them a result, would be able to evolve from their own inner consciousness what the steps were which led up to that result. This power is what I mean when I talk of reasoning backwards, or analytically." A Study in Scarlet
Holmes separates synthetic reasoning, which runs forwards from cause to effect, from analytic reasoning, which runs backwards from an observed state to the history that produced it. Practitioners are trained heavily in the first and barely at all in the second. Yet almost all assessment work is retrodictive. This presentation, this injury, this pattern of missed contacts, what sequence of events produced it.
"It is a capital mistake to theorise before one has data. Insensibly one begins to twist facts to suit theories, instead of theories to suit facts." A Scandal in Bohemia
The canon's clearest statement of what we would now call confirmation bias and anchoring. Holmes states the mechanism accurately, including the important word insensibly. The twisting is not experienced as twisting from the inside.
"One should always look for a possible alternative and provide against it. It is the first rule of criminal investigation." The Adventure of Black Peter
Note the word always. Holmes treats the generation of a competing explanation as a compulsory step rather than a virtue. This maps directly onto the differential diagnosis in medicine, onto the alternative hypothesis in a safeguarding assessment, and onto the debiasing literature, where consider-the-alternative strategies and guided reflection are among the few interventions with any consistent experimental support.
"When you have eliminated the impossible, whatever remains, however improbable, must be the truth." The Sign of Four, and repeated across the canon
Elimination is powerful and it carries a hidden and severe condition. The conclusion is only as good as the completeness of the original list. If the true explanation was never a candidate, elimination will deliver a false answer with maximum confidence, which is the worst of all possible failure modes. Teach the maxim and its condition in the same breath, or do not teach it at all.
"Circumstantial evidence is a very tricky thing. It may seem to point very straight to one thing, but if you shift your own point of view a little, you may find it pointing in an equally uncompromising manner to something entirely different." The Boscombe Valley Mystery
An explicit warning about the seductiveness of a coherent story. Coherence is not validity. A set of facts that fits an explanation neatly will usually fit at least one other explanation just as neatly, and the feeling of neatness is not evidence of anything.
Holmes measures. He times the journey, he examines the ash under the lens, he tries the window, he goes to the place. He almost never accepts a description where an inspection is possible. In The Speckled Band he sits in the room. In The Adventure of the Norwood Builder he raises smoke to flush a hidden space.
The professional translation is the difference between reading a report about a home and standing in it, and the difference between a mother's account of a bedroom and seeing the bedroom. Second-hand information degrades at every transfer, and much of the error described in multi-agency reviews is error introduced by transmission rather than by original observation.
Holmes speaks differently to a duke, a groom, a landlady and a street child. He is not being ingratiating, he is removing the obstacle between the informant and the information. Compare the cognitive interview, and compare motivational interviewing, both of which formalise the same insight. The interviewer's manner is a variable in the quality of the data.
Holmes routinely works on the assumption that his client has told him a partial and self-serving story, without concluding that the client is therefore an enemy. Mary Sutherland in A Case of Identity does not lie, she simply cannot see. Grant Munro in The Yellow Face conceals from shame. This distinction, between the informant who deceives and the informant who cannot or dare not say, is the most useful thing the canon offers to safeguarding practice.
Holmes lets people talk first and questions afterwards. He interrupts rarely during a first account and returns for detail once the free narrative is exhausted. This is exactly the free-recall-then-probe structure that forensic interviewing later established on experimental grounds, and it is a discipline that busy practitioners abandon under time pressure, usually without noticing.
Holmes keeps commonplace books, an index of biographies, files of press cuttings, and he consults them constantly. His famous memory is in large part a filing system. The brain-attic passage in A Study in Scarlet is often quoted as an argument for ignorance, which it is not. It is an argument for deliberate curation of what you carry and deliberate offloading of the rest.
Most multi-agency failure is not a failure of perception by any single professional. It is a failure of integration across professionals and across time. Each individual saw a fragment. Nobody held the whole. Holmes's index is the answer to a problem that our information systems have made worse rather than better.
Holmes establishes sequence obsessively. What happened, in what order, at what hour. He does this before he interprets, not after. A chronology is not an administrative chore that precedes the analysis. Building it is the analysis, because pattern in human affairs is very largely pattern in time. The escalation that becomes obvious when nine agencies put their contacts on one line was invisible to all nine separately.
"I am lost without my Boswell." A Scandal in Bohemia
Holmes needs Watson, and the reason is not companionship. Watson forces articulation. A reasoning process that has to be spoken aloud to an intelligent non-specialist is a reasoning process whose gaps become audible, to the speaker first of all. Every profession has rediscovered this. Rubber-duck debugging in software, the two-challenge rule in aviation, reflective supervision in social work, the ward round presentation in medicine.
Watson also performs a second function which the stories treat as comedy and which is anything but. He is wrong in ordinary ways, and his wrongness marks the path a reasonable person would take. Knowing the attractive wrong answer is a large part of knowing why the right one is right.
Holmes distinguishes what he knows from what he suspects, and he says which is which. "I have devised seven separate explanations, each of which would cover the facts as far as we know them." Compare the practitioner who records a suspicion as a fact and thereby makes it permanent, since a case file has no mechanism for demoting an assertion once written.
Holmes exercises discretion constantly and not always defensibly. He releases James Ryder in The Blue Carbuncle, conceals a killing in The Abbey Grange, and commits burglary in Charles Augustus Milverton. Whatever one makes of the ethics, the stories are unusually rich on the question every practitioner faces weekly. At what point does what I now believe oblige me to act, and what action does it oblige.
"You have erred, perhaps, in attempting to put colour and life into each of your statements instead of confining yourself to the task of placing upon record that severe reasoning from cause to effect which is really the only notable feature about the thing." The Adventure of the Copper Beeches
Holmes is complaining about Watson's prose, and the complaint is the whole of this competency. A record whose merit is atmosphere cannot be audited. A record that sets out the reasoning from cause to effect can be.
The reveal is not vanity. Holmes lays out a chain of inference in order, links it to the observations that support each step, and does so in language the listener can follow. This is precisely what a defensible assessment, a court report or a discharge summary is supposed to do and frequently does not. The most defensible record is not the one with the most confident conclusion, it is the one that shows its working.
"Detection is, or ought to be, an exact science, and should be treated in the same cold and unemotional manner." The Sign of Four
Here the canon is at its least useful and must be taught as a negative example. Holmes treats feeling as contamination. In health and social care the practitioner's own affective response is itself evidence, often the earliest evidence available. The unease on the doorstep, the reluctance to make the next visit, the disproportionate warmth toward one parent. Rather than being cleared away, these need naming, recording and interrogating. This is the one competency where the programme teaches the opposite of what Holmes says, while noting that Holmes in practice is a good deal more moved by people than his manifesto admits.
| Family | Competency | Taught in | Assessed by |
|---|---|---|---|
| A. Perception | A1 Directed attention | M1 | Simulation, observation protocol |
| A2 Baseline knowledge | M2 | Portfolio, normative library | |
| A3 Negative evidence | M5 | Case audit task | |
| A4 Reading traces | M3 | Simulation | |
| B. Inference | B1 Abduction | M4 | Written reasoning trace |
| B2 Reasoning backwards | M4 | Written reasoning trace | |
| B3 Data before theory | M7 | Reflective account | |
| B4 Multiplying alternatives | M7 | Alternatives worksheet | |
| B5 Eliminative reasoning | M7 | Written reasoning trace | |
| B6 Shifting the point of view | M7, M10, M12 | Reflective account | |
| B7 Testing rather than assuming | M3, M9 | Workplace task | |
| C. Elicitation | C1 Adapting register | M6 | Recorded role play |
| C2 Assuming withholding | M6 | Recorded role play | |
| C3 Question sequencing | M6 | Recorded role play | |
| D. Organisation | D1 Externalised memory | M8 | Portfolio index |
| D2 Chronology | M8 | Submitted chronology | |
| D3 Corrective interlocutor | M9 | Peer challenge log, capstone station 3 | |
| E. Judgement | E1 Calibrating confidence | M10 | Capstone |
| E2 Thresholds for action | M10 | Capstone | |
| E3 Communicating reasoning | M11 | Written report | |
| E4 Affect as evidence | M9, M12 | Reflective account |
The structure is blended and distributed rather than intensive, and the reason is that perceptual skill does not transfer from a classroom without rehearsal in the setting where it will be used.
A three-day intensive would be cheaper and it would be worse. What the debiasing literature shows, insofar as it shows anything reliable, is that guided reflection and consider-the-alternative strategies produce the most consistent gains, and that almost nothing has been demonstrated to survive into real clinical settings over the long term. That finding should govern the design. Whatever we teach, we teach with immediate application to live cases, we repeat it, and we build a habit rather than deliver an insight.
Hence thirteen half-days at fortnightly intervals across roughly six months, each followed by a workplace task on the learner's own caseload, evidenced in a portfolio and challenged in a peer group, and then a full assessment day in week 27. Taught contact is 51.5 hours, being thirteen sessions of three and a half hours and a capstone day of six. Guided workplace practice, reading and portfolio work adds a further 38.5, for 90 notional CPD hours.
The programme should not be run as a single-profession course, and the case for mixing is not diversity for its own sake. A nurse and a social worker looking at the same room notice different things, and the discrepancy is the teaching material. Module 9 depends on it entirely. Aim for no more than half the cohort from any one profession, and seat groups so that no table is professionally uniform.
Twelve to eighteen learners. Below twelve the small-group exercises thin out. Above eighteen the observation debriefs cannot be run properly in the time. One lead facilitator throughout, for continuity, with a second facilitator for Modules 6, 12 and 13, and a service user or carer co-facilitator for Modules 6 and 12. The last of these is not optional. Module 12 is about the harm done by professional interpretation, and it cannot credibly be delivered by professionals alone.
On completion the learner will be able to:
| Wk | Module | Family | Workplace task |
|---|---|---|---|
| 1 | M0 You Know My Methods | Induction | Baseline self-audit |
| 3 | M1 You See, But You Do Not Observe | A1 | Protocol trial, three visits |
| 5 | M2 The Brain Attic | A2 | Begin normative library |
| 7 | M3 The Influence of a Trade | A4, B7 | Trace audit of one home |
| 9 | M4 Reasoning Backwards | B1, B2 | Written reasoning trace |
| 11 | M5 The Curious Incident | A3 | Absence audit of one file |
| 13 | M6 A Case of Identity | C1, C2, C3 | Recorded interview |
| 15 | M7 Data Before Theory | B3-B6 | Alternatives worksheet |
| 17 | M8 The Commonplace Books | D1, D2 | Multi-source chronology |
| 19 | M9 Lost Without My Boswell | D3, E4 | Peer challenge log |
| 21 | M10 Circumstantial Evidence | E1, E2 | Threshold case note |
| 23 | M11 The Reveal | E3 | Rewritten assessment |
| 25 | M12 Norbury | Critique | Portfolio completion |
| 27 | M13 Capstone | Assessment | - |
Assessment is by portfolio and by simulation, weighted three to two, and both must be passed. Written examination is deliberately excluded, since it would test recall of the canon rather than the competencies, and would reward exactly the ornamental cleverness the course is trying to discipline.
Assembled across the programme, submitted at week 26. Full requirements in Part Four. Six required items.
Module 13. A live assessment of a constructed scenario with an actor, a physical set and a document bundle, assessed against a published rubric. Learners are marked on process rather than on whether they reach the intended conclusion, and the rubric is deliberately built so that a candidate who reaches the correct conclusion by poor method fails, while a candidate who reaches a defensible wrong conclusion by sound method passes. Say this to the cohort in Module 0 and say it again in Module 12.
Marking process rather than outcome is contestable and the objection deserves an airing with the cohort. Service users are not much comforted by an elegantly reasoned wrong answer. The defence is that outcomes in this work are heavily determined by factors outside the practitioner's control, so outcome marking would score luck, and that over many cases a sound process outperforms a lucky one. That defence is honest but it is not complete, and a good cohort will say so.
The programme has been written at level 7 equivalence and maps to the post-qualifying frameworks the participating professions already use. Local mapping should be completed before the first cohort runs. In England this would ordinarily mean Social Work England CPD requirements, the Nursing and Midwifery Council revalidation portfolio, General Medical Council appraisal and continuing professional development, and Health and Care Professions Council standards for allied health. The portfolio has been designed so that its items serve double duty as revalidation and appraisal evidence, which materially improves completion rates. Check current regulator requirements at the point of delivery, since these change.
Modules 0 to 12 each run three and a half hours including a twenty minute break. Module 13, the capstone, is a full day of six hours. Timings assume a group of fifteen working in five groups of three. Adjust group work upward by four minutes for every additional group.
Induction, baseline and the terms of engagement
"You know my methods, Watson." The Crooked Man, and elsewhere
To establish what the programme is, to take an honest baseline of the cohort's current accuracy and confidence, and to set the ethical terms under which the rest of the course will be taught.
Doyle was Joseph Bell's clerk at the Edinburgh Royal Infirmary in 1877, and wrote to him in 1892 that it was most certainly to him that he owed Sherlock Holmes. The method is a doctor's, borrowed by a novelist, and we are borrowing it back. Open the course with this. It disarms the objection that the material is a literary indulgence before anyone raises it.
| Time | Activity |
|---|---|
| 0.00 | Welcome. The Bell story, five minutes, no slides. |
| 0.05 | Exercise 0.1 The Kit Bag. Cold open, no framing. |
| 0.45 | Reveal and debrief. Introduce the observation and inference distinction from what the group has just done. |
| 1.10 | Programme overview. Structure, portfolio, assessment, expectations of workplace tasks. |
| 1.30 | Break. |
| 1.50 | Exercise 0.2 Baseline self-audit. Individual, silent. |
| 2.20 | Exercise 0.3 The terms of engagement. Whole group, facilitated. |
| 3.00 | The three warnings. Facilitator input, direct and unhedged. |
| 3.15 | Workplace task briefing. Close. |
Time 40 minutes. Grouping Threes.
Preparation. Assemble a shoulder bag belonging to a fictitious person. Suggested contents: a bus pass in a name, a half-finished paperback with a receipt used as a bookmark, a set of keys with a fob from a named gym, a phone charger with a frayed cable, a partly used blister pack of a common medication with two days missing from the middle of the week, a child's drawing folded into quarters, a supermarket receipt, a lanyard with the photograph removed, an unopened letter from a housing association, a packet of paracetamol, hair ties, a pair of reading glasses with one arm taped.
The design principle. Build in three details that support the obvious reading, three that are neutral, and two that contradict it. The bag should invite the conclusion that the owner is a struggling single mother and should in fact belong to a fifty-year-old man caring for his grandchild two days a week. Make the contradicting evidence available but not prominent. The point of the exercise is the moment the group discovers it had the evidence and did not use it.
Instructions to the group. "This bag was handed in. Tell me about the person who owns it. You have twenty five minutes. Write your account on the flip chart, and write next to each statement how confident you are, as a percentage."
Facilitator conduct. Say nothing else. Do not mention observation or inference. Do not warn them. Circulate silently. Note the moment each group commits to a story, and note the time. Most groups commit within the first six minutes and spend the remaining nineteen decorating.
Debrief, 15 minutes.
Time 30 minutes. Grouping Individual, silent, not shared.
Learners complete a written audit which they seal and do not submit until week 26, when they complete it again and write on the difference. Items:
Facilitator note. Item 6 usually produces a number between zero and five per cent and learners find this uncomfortable. Do not soften it. Do not collect the forms.
Time 40 minutes. Grouping Whole group.
The cohort drafts and signs a one-page agreement. Seed it with four propositions and let them argue, amend and add.
Facilitator note. Proposition three is the one that matters and the one groups resist, usually by joking. The material in this course is genuinely pleasurable and the pleasure is the risk. A cohort that has agreed in week one to police its own delight has given you the authority to intervene in week seven without it feeling like a rebuke.
Ask the group a question and leave it open. If Holmes were real, and worked in your service, and carried a caseload of thirty two, would he be any good. Take answers. Do not resolve it. Write it on a sheet and pin it to the wall for the whole programme. Module 12 takes it down.
Before Module 1, write two short accounts of the same visit. One in your normal recording style, one in two columns headed observed and inferred. Bring both.
A Scandal in Bohemia, opening four pages. A Study in Scarlet, chapter two, "The Science of Deduction".
Attention as a resource that must be spent on purpose
"Then how many are there?" "How many? I don't know." "Quite so! You have not observed. And yet you have seen. That is just my point." A Scandal in Bohemia
To convert observation from a virtue into a procedure, and to have every learner leave with a written sweep suited to their own setting.
Drew, Võ and Wolfe asked expert radiologists to search lung scans for nodules and inserted an image of a gorilla, forty eight times the size of the average nodule, into the final case. Eighty three per cent did not report it, and eye tracking showed that most of them had looked directly at it. Expertise does not protect against this. Expertise causes it, because expertise is precisely the narrowing of search. Pair this with Simons and Chabris on the invisible gorilla in the basketball video, and with the fine art observation training used at Yale, where medical students taught to describe paintings in a gallery improved their accuracy in describing clinical photographs. The instruction that worked was not "look harder". It was "describe everything before you interpret anything".
| Time | Activity |
|---|---|
| 0.00 | Exercise 1.1 Seventeen steps. Learners leave the room and return. |
| 0.20 | Input. Seeing against observing. Attention as allocation. The gorilla studies. |
| 0.45 | Exercise 1.2 Building the sweep. Small groups by profession. |
| 1.25 | Break. |
| 1.45 | Exercise 1.3 The Baker Street Room. Two conditions, run in halves. |
| 2.45 | Scored debrief. Yield comparison. |
| 3.10 | Counterpoint. The cost of the protocol. |
| 3.25 | Workplace task. Close. |
Time 20 minutes. Grouping Individual then pairs.
Before any input, ask the group to write down, without leaving their seats, the answers to six questions about the building they have all walked through. How many steps from the entrance to this room. What colour is the front door. Is there a fire extinguisher in the corridor and where. What is written on the notice nearest the lift. How many chairs are in this room. What is directly behind you.
Then send them to check. Ten minutes. They return with the answers.
Debrief. The point is not that they failed, since everyone fails. The point is the second question you ask. "Which of those six could you have answered if you had known in advance you would be asked?" All of them. Attention follows expectation, so the practitioner who knows what they are going to look for will see it and the practitioner who intends to be generally observant will not.
Time 40 minutes. Grouping Threes, grouped by profession for this exercise only.
Each group writes a first-visit observation protocol for their own setting on a single card. Constraints, given in this order and not before:
Facilitator note. The second constraint is the whole exercise and groups will violate it repeatedly. Every time a card says "signs of", send it back. Practitioners are so habituated to outcome language that they find it genuinely difficult to write down a place rather than a judgement, and watching them struggle with it is more instructive than being told about it.
Cross-profession swap. Last ten minutes. Each group passes its card to a group from a different profession, who must add one item and delete one. Defend both.
Time 60 minutes running, 25 debrief. Grouping Two halves.
Preparation. Dress a room as the living room of a fictitious household. Plant approximately thirty five details across four categories.
Procedure. Half the cohort enters first with no protocol and eight minutes. They write what they saw. The second half enters with their Module 1 protocol cards and the same eight minutes. Neither half sees the other's list until the debrief. Then swap conditions with a second, differently dressed area if you have one, so that nobody spends the whole exercise in the disadvantaged condition.
Scoring. Score each half on total items, on subtle items and on absences. The interesting result is almost always the same. Protocol users find slightly more items overall, considerably more subtle items and dramatically more absences. Show the three numbers separately. If you show only the total, the exercise looks unimpressive and the lesson is lost.
Debrief questions.
A checklist directs attention, and directed attention is attention withdrawn from everywhere else. The gorilla was missed by radiologists who were searching properly, and searching properly is what caused the miss. A protocol therefore buys reliability at the price of a particular kind of blindness, and the trade is worth making only if two conditions hold. The protocol must be short enough to leave residual capacity, and it must end with an unstructured item. The last line of every card should read: now look again at whatever you have not looked at.
Put the objection to the group before you resolve it. Ask what a protocol would have to do to catch the thing it was not written for. Somebody usually reinvents the free sweep, which is a better outcome than being told.
Use the protocol on three visits. Record yield against your normal practice. Note one occasion where it helped and one where it got in the way. Both go in the portfolio.
A Scandal in Bohemia. Drew, Võ and Wolfe on inattentional blindness in expert observers. Simons and Chabris on the invisible gorilla.
Baselines, normality, and curiosity as a knowledge base
"I consider that a man's brain originally is like a little empty attic, and you have to stock it with such furniture as you choose... It is a mistake to think that that little room has elastic walls and can distend to any extent." A Study in Scarlet
To replace the exhortation to be professionally curious with the construction of a normative library, and to establish that anomaly detection depends on prior knowledge of the typical.
De Groot found that chess masters reconstruct a board from five seconds' exposure far better than weaker players. Chase and Simon later added the control that matters. The advantage holds only where the position is one that could arise in a real game, and on randomly scattered pieces it collapses. Expertise is not superior memory, it is a library of typical patterns. Now put that beside Kahneman and Klein's joint paper of 2009, in which two researchers who had spent careers disagreeing about intuition agreed on when it works. Intuition is trustworthy where the environment is sufficiently regular to be predictable and where the practitioner has had the opportunity to learn those regularities through prolonged practice with feedback. Both conditions. Feedback is the one that fails in social care, because the practitioner very often never learns what happened. That is not a small caveat, it is the central problem of the profession, and it deserves twenty minutes of the group's time.
| Time | Activity |
|---|---|
| 0.00 | Return of workplace task. Protocol findings, plenary. |
| 0.20 | Input. The brain attic properly read. Anomaly requires a norm. |
| 0.40 | Exercise 2.1 What do you know the normal range of. |
| 1.20 | Break. |
| 1.40 | Exercise 2.2 The base rate test. |
| 2.20 | Input and discussion. Conditions for valid intuition. Where feedback fails. |
| 2.45 | Exercise 2.3 Designing your feedback loop. |
| 3.15 | Counterpoint. Whose normal. Close. |
Time 40 minutes. Grouping Individual then mixed threes.
Each learner writes two lists. Ten things in their working world where they have a confident sense of the normal range, and ten where they know they do not. Fifteen minutes.
Then, in mixed-profession threes, each learner teaches the other two one item from their first list and asks for teaching on one item from their second. Twenty minutes. Five minutes to write down what they learned.
Facilitator note. This exercise works because the second list is far harder to write than the first, and because the mixing exposes how narrow each profession's library is. A district nurse knows the normal range of a pressure area, a housing officer knows the normal range of rent arrears behaviour, a social worker knows the normal range of a contact record. None of them knows the others', and cases live in the overlap.
Prompts for a stuck group. How much post accumulates in a fortnight. What does a fridge look like at the end of the month rather than the start. How long does a bruise on a shin take to change colour. What proportion of appointments does an average patient miss. How many pill boxes are in an ordinary kitchen cupboard. What time do children in your area go to bed.
Time 40 minutes. Grouping Individual, then plenary.
Twenty questions with numerical answers, drawn from the group's actual operating environment and verified in advance from local data. Learners write an estimate and a range they are ninety per cent confident contains the true value.
Example items, to be localised: what proportion of children on a child protection plan in this local authority are on it for neglect. What proportion of referrals to your service are closed at first assessment. What proportion of adults over eighty five in this area live alone. How many households in this ward are in temporary accommodation. What proportion of patients presenting with chest pain in your setting have a cardiac cause.
Scoring. Count how many of the twenty true values fell inside each learner's ninety per cent range. A well calibrated person scores eighteen. Most professionals score between eight and twelve, which means their ninety per cent ranges are functioning as fifty per cent ranges.
Debrief. The lesson is not that people are ignorant of statistics. It is that professional judgement operates on a stream of cases that is not representative of the population, and that we quietly substitute our caseload for the world. The social worker who sees neglect all day will overestimate its prevalence, and will therefore see it in an ambiguous case. Holmes never mentions a base rate anywhere in the canon, which is one of the clearest places where the fictional method departs from a sound one.
Time 30 minutes. Grouping Pairs.
Each learner identifies one recurring judgement they make where they currently never find out whether they were right. They then design a practicable method of finding out. It must be specific, must not depend on anyone else's goodwill, and must be achievable within the six months of the course.
Examples that have worked. A GP who logged every case where she was uncertain and reviewed the notes at three months. A duty social worker who recorded his predicted outcome on every screening decision and audited fifty of them a quarter later. A discharge coordinator who rang twenty patients at two weeks.
Facilitator note. Expect resistance on grounds of time, and expect it to be sincere. Concede the point and then press. An hour a quarter is the price of knowing whether twenty years of experience is twenty years of learning or one year repeated twenty times.
A baseline is a description of what is usual in a population, and the practitioner's sense of usual is built from the population they have happened to encounter. That makes it a cultural artefact wearing the clothes of an observation. The tidy house, the child in bed by eight, the parent who makes eye contact, the relative who asks questions in a meeting. Every one of these is a class marker and a cultural marker before it is anything else, and reading them as indicators of care is one of the mechanisms by which services discriminate while believing themselves to be observing.
Ask the group directly. Name one thing you treat as a warning sign that is really a marker of class. Give them a minute of silence first. The answers are usually good and they are usually uncomfortable.
Begin the normative library. A single document, one page per domain, recording the normal range you have observed and the source. Add to it for the rest of the programme. Implement the feedback loop designed in Exercise 2.3.
A Study in Scarlet, the brain attic passage. Kahneman and Klein, "Conditions for intuitive expertise: a failure to disagree", 2009. De Groot on chess reconstruction, with Chase and Simon on random positions.
Reading traces in objects, environments and bodies
"By a man's finger-nails, by his coat-sleeve, by his boot, by his trouser-knees, by the callosities of his forefinger and thumb, by his expression, by his shirt-cuffs, by each of these things a man's calling is plainly revealed." A Study in Scarlet
To teach the disciplined reading of physical evidence, and in the same session to teach the limits of it, since this is the material that most readily degrades into confident prejudice.
Forensic science has spent twenty years discovering that Holmes was optimistic about it. Bite mark comparison and microscopic hair comparison have both been substantially discredited, and in 2004 the FBI matched a latent fingerprint from the Madrid train bombings to Brandon Mayfield, an Oregon lawyer who had never been to Spain. Three examiners concurred. Itiel Dror's subsequent work showed that fingerprint examiners shown the same prints twice, with different contextual information attached, changed their own conclusions. The examiners were not incompetent and they were not corrupt. They were told a story first, and the story shaped what the ridges looked like. Every practitioner in the room reads a referral before they meet the person, and that referral does to them exactly what the context did to the examiners. Say this before the object exercises rather than after.
| Time | Activity |
|---|---|
| 0.00 | Normative library check-in. Three volunteers, five minutes each. |
| 0.20 | Input. The forensic reliability problem and contextual bias. |
| 0.45 | Exercise 3.1 Prosecuting the hat. |
| 1.25 | Break. |
| 1.45 | Exercise 3.2 The object table. |
| 2.45 | Exercise 3.3 Contextual contamination. |
| 3.15 | Debrief and close. |
Time 40 minutes. Grouping Fours, in role.
Distribute the passage from The Blue Carbuncle in which Holmes reads Henry Baker's hat. He concludes that the owner was intellectual, that he was formerly well-to-do and has fallen on hard times, that he has taken to drink, that his wife has ceased to love him, that he leads a sedentary life, that he goes out little, that he is out of training, and that his house is not lit by gas.
Each group of four takes one inference and splits into two pairs. One pair prosecutes, that is, states the evidence and the reasoning. The other pair defends, that is, produces at least two alternative explanations for the same physical evidence and identifies what further information would discriminate between them. Twenty five minutes. Then each group reports in ninety seconds.
Facilitator note. The chain to steer people toward is the wife. Holmes observes dust on the hat and infers that no one has brushed it, and from that infers that the owner's wife has ceased to love him. Watson does not object. The group will find at least six alternatives within a minute, and the exercise is over the moment they realise that the most famous inference in English fiction rests on an assumption about who brushes hats in 1890. That is the lesson. A chain of inference is only as sound as its least examined social assumption, and the least examined assumptions are always the ones the practitioner shares with everyone in the room.
Debrief question. What is the assumption in your own assessment framework that is doing the same work as "wives brush hats".
Time 60 minutes. Grouping Threes, rotating.
Preparation. Eight objects, each with a genuine history you can reveal afterwards. Suggested set: a worn shoe with uneven sole wear, a weekly medication organiser filled to Thursday, a child's winter coat with the cuffs let down, an electric kettle with heavy limescale on one side, a walking frame with worn grips on the left handle only, a mobile phone with a cracked screen and a charging port full of lint, a photograph of the inside of a fridge, a bank statement with the names redacted.
Groups rotate through the objects at six minutes each. For each object they complete three columns.
Rule. No entry in column two is permitted without an entry in column three. This is the discipline the whole course is built on and it starts here.
Reveal. Ten minutes at the end. Read out the true history of each object. Design at least three of them so that the obvious reading is wrong. The frame with one worn grip belongs to someone with a shoulder injury rather than a stroke. The medication organiser was filled by a daughter who visits on Thursdays and not by a patient who stopped taking her tablets.
Time 30 minutes. Grouping Two halves, blind to each other.
Procedure. Both halves receive the same photograph of a living room and the same short referral. The referrals differ in one respect only. Half are told the referral came from a neighbour concerned about shouting. Half are told the referral came from a district nurse concerned that the client is isolated since her husband died.
Each learner writes six observations and one working conclusion. Ten minutes, silent, no discussion.
Then the halves compare. In every run of this exercise the two halves have noticed different things in the same photograph, and both halves believe they simply looked at what was there.
Debrief. Return to Dror's examiners. Then ask the practical question. Given that you cannot work without reading the referral, what can you actually do. Answers worth having include reading the referral after the first observation rather than before, recording your observations before your conclusion, and having a colleague who has not read the referral look at the same material. Note that all three are procedural rather than attitudinal. You cannot decide not to be biased.
Every trace-reading skill taught this morning can be turned into an instrument for penalising people for being poor. A cold house, an empty fridge, a broken washing machine, a child in clothes too small. Each is a physical trace and each has at least two explanations, one about care and one about money, and the professions have a long record of choosing the first. Make the group state the discriminatory version of each object they read this morning. If the exercise ends with the room pleased with itself, the module has failed.
Conduct one visit using the three-column record. Submit it with every column completed. If any inference lacks an alternative, the task is returned.
The Blue Carbuncle, the hat. A Study in Scarlet, the Book of Life passage. Dror on contextual bias in fingerprint examination, and the Office of the Inspector General report on the Brandon Mayfield case, 2006.
Abduction, retrodiction and the analytic method
"There are few people, however, who, if you told them a result, would be able to evolve from their own inner consciousness what the steps were which led up to that result. This power is what I mean when I talk of reasoning backwards, or analytically." A Study in Scarlet
To name the form of reasoning the profession actually uses, to distinguish it from the form it thinks it uses, and to give learners a written structure that makes their reasoning auditable.
Whole sciences are retrodictive and have had to build their own safeguards. Geology, palaeontology and evolutionary biology cannot run the experiment, so they reason from present traces to past causes, which is exactly what an assessment does. Cuvier claimed to reconstruct an animal from a single bone, and was right often enough to become famous and wrong often enough to be instructive. John Snow reasoned backwards from the geography of cholera deaths to the Broad Street pump without knowing what cholera was, which is the best single demonstration available that you can reach a correct and actionable conclusion about a mechanism you do not understand. Air accident investigation is the closest institutional cousin to this work, and it is worth showing the group an Air Accidents Investigation Branch report for the discipline of its sequencing and for its refusal to name a single cause.
| Time | Activity |
|---|---|
| 0.00 | Return of three-column records. Plenary on alternatives that were hard to find. |
| 0.20 | Exercise 4.1 Sorting the inferences. |
| 0.50 | Input. Why Holmes is not deducing, and why that matters. |
| 1.15 | Break. |
| 1.35 | Exercise 4.2 The retrodiction drill. |
| 2.25 | Exercise 4.3 Writing the reasoning trace. |
| 3.10 | Counterpoint. Underdetermination. Close. |
Time 30 minutes. Grouping Threes.
Give each group twenty statements on cards, a mixture drawn from the canon and from real anonymised case records. They sort into deduction, induction and abduction, and then into a fourth pile for anything that is none of the three.
Sample cards. "All patients on this ward have been screened, Mr A is on this ward, therefore Mr A has been screened." "Every child I have seen with these marks had been struck, therefore children with these marks have been struck." "The bruise is on the ear, the ear is a site rarely injured accidentally, therefore this was probably inflicted." "Watson is tanned and his arm is stiff, therefore he has come from Afghanistan." "Mother appeared evasive, therefore she is concealing something."
Debrief. Almost everything in a real case file is abduction, and almost all of it is written in the grammar of deduction. The word therefore is doing work it has not earned. Ask the group to find one therefore in their own recent recording that should have been a which would be consistent with.
The last card is the important one. "Mother appeared evasive, therefore she is concealing something" contains an unexamined observation as well as an unwarranted inference, and Module 6 will take it apart.
Time 50 minutes. Grouping Threes.
Preparation. A photographic end state with a known history. A kitchen at nine in the morning. The group is told only what is in the photograph and is asked to reconstruct the previous fourteen hours.
Procedure in three rounds.
Debrief, 10 minutes. Round three is the transferable skill and the one nobody teaches. The practitioner's scarce resource is not thought, it is enquiry, and knowing which single question would move you furthest is worth more than any amount of speculation. Holmes does this constantly. He does not gather everything, he gathers the one thing that decides between his candidate explanations.
Time 45 minutes. Grouping Individual then pairs.
Introduce the programme's standard reasoning trace, which learners will use for the rest of the course and submit in the portfolio. Six fields on one page.
| Field | Content |
|---|---|
| 1. Observed | What I saw, heard or read. Sources named. No interpretation. |
| 2. Best explanation | The account that currently fits best, stated in one sentence. |
| 3. Competing explanations | At least two. Each must be one a reasonable colleague could hold. |
| 4. Discriminating information | What would tell these apart, ranked by how obtainable it is. |
| 5. Confidence | A verbal statement matched to the programme's confidence scale, with the reason for that level. |
| 6. Trigger to revise | What would have to happen for me to abandon field two. Written before the outcome is known. |
Learners complete one on a live case from their own work. Then swap with a partner from another profession, who must attack field three. If a competing explanation is a straw man, it is sent back.
Facilitator note. Field six is the innovation and the one to defend hardest. A working hypothesis with no stated defeater is not a hypothesis, it is a belief. Writing the defeater down in advance is the only reliable protection against the phenomenon in which a case narrative absorbs all subsequent information and never changes.
Underdetermination is the formal name for the objection and the group needs it. Any finite set of present traces is compatible with an infinite number of past histories, so retrodiction can never be conclusive and can only ever be comparative. Holmes escapes this because a novelist constructed the traces to have exactly one plausible parent. Real evidence is not curated. It is the residue of a world with no author, and it will often be compatible with a dozen histories none of which is more elegant than the others.
There is a sharper version worth putting to a confident group. Serious case reviews are also written backwards, by people who know the outcome, and they routinely produce a narrative in which the warning signs were legible all along. That is the Holmes fallacy performed by the very process designed to correct it. Hindsight manufactures coherence. If your review reads like a detective story, be suspicious of it.
Complete two reasoning traces on live cases. Field six must be written before the outcome is known. Bring both.
A Study in Scarlet, chapter two. Peirce on abduction, any short introduction. One Air Accidents Investigation Branch report of the facilitator's choosing, read for its sequencing rather than its subject.
Negative evidence and the things that did not happen
"Is there any point to which you would wish to draw my attention?" "To the curious incident of the dog in the night-time." "The dog did nothing in the night-time." "That was the curious incident." Silver Blaze
To make absence perceptible, which requires a procedure, because ordinary attention registers events and not the failure of events.
Abraham Wald was asked during the Second World War where to add armour to bombers, and was shown the distribution of bullet holes in the aircraft that came back. The obvious answer was to armour where the holes were. Wald's answer was the opposite. The returning planes were the survivors, so the undamaged areas were where a hit was fatal. The data was a record of absence and everyone had read it as a record of presence. Ask the group where the equivalent survivorship data sits in their own service. The families who stop engaging. The patients who do not come back. The referrals that were never made. Every service knows a great deal about the people it sees and almost nothing about the people it has lost, and it draws conclusions from the first as though they described the second.
| Time | Activity |
|---|---|
| 0.00 | Reasoning traces returned. Focus on field six. |
| 0.20 | Read Silver Blaze extract aloud. Input on negative evidence. |
| 0.40 | Exercise 5.1 The file with holes. |
| 1.30 | Break. |
| 1.50 | Exercise 5.2 The absence taxonomy. |
| 2.30 | Exercise 5.3 Who have you never seen. |
| 3.05 | Counterpoint. Over-reading silence. Close. |
Time 50 minutes. Grouping Threes.
Preparation. A constructed case bundle of about forty pages covering eighteen months. Multi-agency, realistic in its dullness. Build in eleven planted absences across four types.
Instruction. "Do not tell me what happened. Tell me what is not here." Thirty five minutes reading and listing. Fifteen minutes plenary.
Facilitator note. Groups will still try to construct a narrative, because that is what a file invites. Interrupt at the fifteen minute mark and ask each group to read out its list so far. Any group that has produced a story rather than a list of gaps should be sent back with the instruction repeated. The discomfort is the point. Reading for absence is an unnatural act.
The one that always gets missed. Plant a child who is named in a single line of a health visitor record in month two and never appears again. Every cohort misses this and every cohort recognises the pattern from published reviews the moment it is pointed out.
Time 40 minutes. Grouping Mixed threes then plenary.
The group builds a working taxonomy of absences that matter in their settings, and for each type specifies where you would have to look to detect it. The taxonomy below is a starting point and the group should extend it.
| Type | Example | Where you would look |
|---|---|---|
| Person never seen | A resident adult who has never been present at a visit | Cross-reference visit records against household composition |
| Voice never heard | The person's own account absent from an assessment about them | Search the record for direct quotation |
| Contact not made | A pattern of non-attendance treated as administrative | Attendance data rather than case notes |
| Escalation not occurring | A condition that should have progressed and has not | Chronology of measurements over time |
| Question never asked | No record of asking about a domain the framework requires | Assessment against framework headings |
| Agency never involved | An obvious service that has no footprint in the file | List of expected agencies against those present |
Key teaching point. Every row's third column is a systems query rather than an act of perception. Absence is generally not detectable by looking harder at the case. It is detectable by comparing the case against an expectation, which means the expectation has to be written down somewhere first. This is where the normative library from Module 2 does its real work.
Time 35 minutes. Grouping Individual, silent, then pairs.
Each learner takes their own current caseload and answers four questions in writing.
Facilitator note. Question four is the one that matters and it must be introduced without judgement, because the honest answer is universal and admitting it is not. Avoidance is data about the case and it is usually data about risk. This anticipates Module 9, where affect is treated as evidence rather than as noise. Do not collect the answers. Do ask, at the end, for a show of hands on whether anyone intends to do something differently this week.
The maxim that absence of evidence is not evidence of absence is usually quoted to defend inaction, and it is half right. An absence in a record is often a fact about the record. Missing documents mean an administrator was off sick more often than they mean a family is concealing something, and a service that trains its staff to treat every gap as suspicious will generate a great deal of intrusion and a great deal of harm.
The test worth teaching is conditional. An absence carries information only where its presence was expected. The dog is evidence because a dog would have barked at a stranger. Four missed appointments carry information only if you know the attendance base rate for that clinic, which returns the group to Module 2. Ask the group to state, for three absences they found this morning, what expectation makes each one meaningful. Some will survive the test and some will not.
Conduct a full absence audit on one open case using the taxonomy. Record what you found and what action followed. This item is required for the portfolio.
Silver Blaze. An account of Wald's work on aircraft armour and survivorship. One published review from the learners' own sector, read for what is absent from the record rather than for what happened.
Elicitation, testimony and the informant who cannot say
"It has long been an axiom of mine that the little things are infinitely the most important." A Case of Identity
To teach a structured elicitation sequence, and to replace the crude opposition between truth and lying with a working account of why people do not say what they know.
Loftus showed that witnesses asked how fast cars were going when they smashed into each other gave higher estimates than those asked when the cars hit each other, and were later more likely to report broken glass that was never there. The question does not extract the memory, it participates in building it. Fisher and Geiselman's cognitive interview and the Achieving Best Evidence guidance both descend from that finding, and both rest on the same sequence Holmes uses without naming it. Let the account run uninterrupted first, probe afterwards, and never supply the detail you are hoping to hear. Set this beside motivational interviewing, which arrives at a similar discipline from an entirely different direction, out of addiction treatment rather than forensic psychology. Two fields, no contact, same conclusion. That convergence is worth pointing out to the group.
This module requires a second facilitator and two trained actors, and it should include a service user or carer co-facilitator for the final debrief. Do not run it with learners role playing service users. They perform their own assumptions and the exercise teaches the wrong thing.
| Time | Activity |
|---|---|
| 0.00 | Absence audits returned. Ten minutes plenary. |
| 0.15 | Input. Free recall then probe. The Loftus problem. |
| 0.40 | Exercise 6.1 The contaminating question. |
| 1.10 | Break. |
| 1.30 | Exercise 6.2 The four reasons people do not tell you. |
| 2.05 | Exercise 6.3 Recorded interview with actor. |
| 3.00 | Debrief with service user co-facilitator. |
| 3.25 | Close. |
Time 30 minutes. Grouping Pairs then plenary.
Distribute a transcript of a real, anonymised assessment interview of about four pages. Pairs mark every question as open, closed, leading, compound or contaminating, and count them. Then they rewrite the worst six.
Categories to define first. A leading question supplies the answer. A compound question asks two things and gets one answer, and you will not know which. A contaminating question introduces a fact the informant had not supplied, after which you can never establish whether they knew it independently.
Facilitator note. Use a transcript from the learners' own sector and make sure it is a competent interview rather than an obviously bad one. The lesson is that ordinary, professional, well-intentioned interviewing is full of contamination, and that a transcript of the group's own work would look the same. Say so.
Time 35 minutes. Grouping Mixed fours.
Introduce the taxonomy, then have groups populate it from their own experience and specify a response for each. This is the module's core content and it should be given to the group as a distributed card, since practitioners use it long after the course ends.
| Type | What it looks like | What helps | What makes it worse |
|---|---|---|---|
| Cannot say Does not have the information | Vagueness, deference, agreeing with whatever is suggested | Ask about routines rather than events. Ask what usually happens | Persistence, which produces compliance rather than information |
| Does not know it matters | Full and open account that omits the relevant thing entirely | Systematic coverage of domains. The framework as a checklist of topics not questions | Assuming that an open account is a complete one |
| Dare not say Fear, shame, loyalty, consequence | Changes of subject, minimising, humour, sudden practical concerns | Naming the cost of telling. Time alone. A second visit. Explicit statement of what you will and will not do with the information | Reassurance you cannot honour. Recording it as evasiveness |
| Will not say Chooses not to | Consistent, controlled, competent refusal | Respecting it, recording it accurately, being clear about consequences | Treating a decision as a symptom |
The teaching point. Mary Sutherland in A Case of Identity is not lying and is not concealing. She cannot see, and she cannot see because seeing would cost her more than she can pay. Holmes works this out and, notably, decides not to tell her. Whether he was right is a question worth ten minutes, and it leads directly into Module 10.
Practitioner reflection. Ask each learner to recall a case where they recorded someone as evasive or resistant. Which of the four was it actually. Most of the room will find they used a category-four word for a category-three situation, and the record will have hardened accordingly.
Time 55 minutes. Grouping Rotating threes with an actor.
Procedure. Two actors work with briefs written so that each holds information in a different category from the taxonomy. One dare not say. One does not know it matters. Learners rotate in threes, one interviewing, one observing question type, one keeping time and marking the transition from free recall to probing. Twelve minutes per interview, three rotations.
Mandatory structure.
Facilitator note. The four minutes of silence is extremely hard and almost nobody manages it first time. Record the violation count and show the group the distribution. The average practitioner interrupts a free account within about forty seconds, and almost always to ask something the informant was about to say.
Debrief with the service user co-facilitator, 25 minutes. This is the most valuable twenty five minutes in the programme. The prompt is simple. What did that feel like from the other chair.
Holmes obtains information by disguise, by false pretences and in Charles Augustus Milverton by burglary and a feigned engagement to a housemaid. Watson raises the ethical objection and Holmes bats it away. No practitioner in the room may do any of this, and the professional codes are right about that.
The deeper objection is not about deception. Holmes's informants are sources, and information flows one way. Contemporary practice in both health and social care is built on the opposite premise, that the person is a partner in understanding their own situation and holds an authority over it that the professional does not. A method that treats the person as a text to be read sits badly with that, and the tension is real rather than rhetorical. Put it to the group as a question rather than an answer. When is a service user a source of evidence, and when are they the author of the account, and what do you do when your judgement is that they are wrong about their own life.
Conduct and record one real interview with consent, using the four phase structure. Transcribe five minutes of it. Mark your own question types. Submit with a two hundred word reflection.
A Case of Identity. The Yellow Face, for Grant Munro's concealment. Loftus and Palmer on the wording of questions, 1974. Fisher and Geiselman on the cognitive interview. The current Achieving Best Evidence guidance, the interview structure sections only.
Premature closure and the discipline of alternatives
"It is a capital mistake to theorise before one has data. Insensibly one begins to twist facts to suit theories, instead of theories to suit facts." A Scandal in Bohemia
To demonstrate premature closure in the room rather than describe it, and to install two procedures that have some experimental support against it.
Wason gave people the sequence two, four, six, told them it followed a rule, and invited them to test their guesses by proposing further triples. Most people propose eight, ten, twelve, then twenty, twenty two, twenty four, receive a yes each time and announce with confidence that the rule is ascending even numbers. The rule is any ascending sequence. They never once proposed a triple they expected to be rejected. This is confirmation bias in its purest laboratory form and it runs live in ten minutes with any group. Run it. It humbles a room in a way no lecture can, and the humbling is necessary before anyone will do the tedious work of writing down alternatives they do not believe.
| Time | Activity |
|---|---|
| 0.00 | Exercise 7.1 Two, four, six. Cold open. |
| 0.20 | Input. Anchoring, confirmation, premature closure, the search satisficing pattern. |
| 0.45 | Exercise 7.2 The commitment clock. |
| 1.25 | Break. |
| 1.45 | Exercise 7.3 The pre-mortem. |
| 2.30 | Exercise 7.4 What was never on the list. |
| 3.05 | Counterpoint. The evidence for debiasing is thin. Close. |
Time 20 minutes. Grouping Threes.
Run Wason's task exactly. Give the triple, state that it conforms to a rule you have in mind, invite each group to test triples on you and to announce the rule only when confident. Answer yes or no and nothing else. Keep a public tally of how many triples each group proposed that they expected to be rejected.
Debrief. The tally is usually zero across the room. Then ask the question that transfers it. What is the case on your caseload right now where every enquiry you have made in the past month was one you expected to confirm what you already think.
Time 40 minutes. Grouping Threes.
Procedure. A staged case unfolds in six timed releases, four minutes apart. After each release every learner writes privately their current best explanation and their confidence. They may change it at any release but must log the change.
Design of the releases. Releases one and two support explanation A strongly. Release three is neutral. Release four is mildly inconsistent with A. Release five is strongly inconsistent with A and consistent with B. Release six settles it as B.
The finding. Plot the room's confidence in A across the six releases. Typically confidence rises through releases one to three, holds flat through four, and only a minority move at release five. Show the graph. The interesting group is not those who never moved, it is those who moved at four, and it is worth asking them what they did differently. Frequently the answer is that they had written A down as provisional rather than as a conclusion.
Facilitator note. Run this with private written logs rather than open discussion, or the room converges and the effect disappears. Collect the logs anonymously and plot them in the break.
Time 45 minutes. Grouping Fours, mixed profession.
Each learner brings a decision they are about to make on a live case. In fours, take each in turn, ten minutes each.
The instruction, given exactly. "It is eighteen months from now. This decision has gone badly wrong and there is a review. You are reading the review. Write the paragraph that explains why it went wrong."
The tense matters. Asking what might go wrong produces a list of risks nobody believes. Asking why it did go wrong, in the past tense, with the failure stipulated, produces specific and uncomfortable answers, because the mind is no longer defending the decision but explaining it.
Then. Each learner identifies which of the failure explanations they can act on this week, and writes it into field six of their reasoning trace as a trigger to revise.
Time 35 minutes. Grouping Plenary then threes.
Take the maxim about eliminating the impossible and put it on the wall. Then take three published case reviews and, for each, ask the group to establish what the true explanation was and whether it was ever a candidate at the time.
The point. In most reviewed cases the correct explanation was never eliminated because it was never entertained. Elimination therefore delivered a confident wrong answer, and confidence was highest precisely because the list had been worked through carefully. Rigour applied to an incomplete list produces error with a clean audit trail, which is more dangerous than sloppiness, since sloppiness at least looks like what it is.
Procedure to install. Before eliminating, ask the list question. Who has not contributed to this list. A member of another profession, the person themselves, a family member, someone with no knowledge of the case. Then ask what kind of explanation this team is structurally unlikely to generate. Every team has a blind spot shaped like its own composition.
Be honest with the cohort, because they will find out anyway and the credibility of the whole programme rests on it. Systematic reviews of debiasing in medical diagnosis find that guided reflection has the most consistent support and that consider-the-alternative strategies have some. They also find mixed results across studies, no studies conducted in real clinical settings, and no long-term follow-up. What we know is that these procedures can improve diagnostic accuracy among students and residents under exam conditions. What we do not know is whether they survive contact with a full clinic on a Friday afternoon.
That is a weaker claim than the one this module has spent three hours making, and stating it plainly is the point. A course about the discipline of alternatives that presents its own methods without alternatives has failed its own test. Ask the group what would have to be true for these procedures to work in their setting, and what would have to be true for them to be a waste of time. Both lists are usually short and both are usually about caseload.
Complete an alternatives worksheet on one live case, including a pre-mortem. Bring the case to Module 9 for peer challenge.
A Scandal in Bohemia, the opening. The Adventure of Black Peter, the alternative rule. Wason on the 2-4-6 problem, 1960. Klein on the pre-mortem. One systematic review of debiasing in medical diagnosis, read for its limitations as much as its findings.
Chronology, records and the memory a service does not have
"For years he had adopted a system of docketing all paragraphs concerning men and things, so that it was difficult to name a subject or a person on which he could not at once furnish information." A Scandal in Bohemia
To treat information architecture as a clinical skill rather than an administrative burden, and to establish chronology as the primary analytic instrument of multi-agency work.
Holmes's scrapbooks belong to a tradition older than the detective story. The early modern commonplace book was a working technology for reading, and John Locke published a method of indexing one in 1686. Niklas Luhmann built a card index of ninety thousand slips and credited his output to it rather than to himself, on the grounds that the index could produce combinations he had not thought of. Vannevar Bush described the same idea as a machine in 1945 and called it the memex. What every one of these has in common, and what our case management systems conspicuously lack, is that they were built to be read across rather than stored in. A chronology is the same move. It takes records organised by agency and reorganises them by time, and the reorganisation is where the meaning is. Ask the group how many of their systems allow that view. The answer is usually none, and the workaround is usually a spreadsheet built by hand at three in the morning before a conference.
| Time | Activity |
|---|---|
| 0.00 | Alternatives worksheets returned. Brief plenary. |
| 0.15 | Input. Chronology as analysis, not as preparation for analysis. |
| 0.35 | Exercise 8.1 The chronology build. |
| 1.35 | Break. |
| 1.55 | Exercise 8.2 Reading the chronology. |
| 2.30 | Exercise 8.3 Writing for the stranger. |
| 3.00 | Exercise 8.4 Designing your index. |
| 3.20 | Counterpoint. Records that serve the organisation. Close. |
Time 60 minutes. Grouping Fours.
Preparation. The Module 5 case bundle, extended. Records from six agencies covering two years, deliberately in six different formats and with three different date conventions. Include two documents where the date of the event and the date of recording differ by weeks, since this is the commonest source of a false chronology and it is almost never taught.
Standard to work to. One line per event. Seven columns, as set out in the chronology standard in Part Four.
Facilitator rules, enforced. No group may write in the significance column during this exercise. Groups will want to, and stopping them is the whole design. Building and interpreting at the same time produces a chronology that has been curated to support the interpretation, which is the most common defect in the ones that reach case conferences.
Time 35 minutes. Grouping Same fours, then plenary.
Now the significance column. Groups work through five specific questions rather than reading impressionistically.
Facilitator note. Question five is the one that surprises people. A widening recording lag is usually a signal about the professional or the service rather than the family, and it is often the earliest available indicator that a case is being avoided. Link back to Module 5, question four.
Plenary. Ask each group for one thing the chronology showed that forty pages of narrative did not. There is always at least one, and hearing four of them makes the argument better than any input could.
Time 30 minutes. Grouping Pairs.
Each learner brings a case note they wrote in the past month. They rewrite it to a single standard. A colleague from another agency, picking this up in three years, with no knowledge of the case and no ability to ask you, must be able to establish what happened, who said it, what you concluded and why.
Common defects to name in advance. Pronouns without antecedents. Initials. Local acronyms. "As discussed." "Usual concerns." Conclusions with no attached evidence. Evidence with no attached source. The passive voice concealing who did something. Dates written as "last week".
Test. Partners read each other's rewritten note cold and attempt to answer the four questions. Where they cannot, the note is not finished.
Time 20 minutes. Grouping Individual.
Each learner designs a personal index, being the thing they will maintain outside the corporate system and take with them between jobs. It should hold the normative library from Module 2, the base rates from their own audits, the local knowledge that currently lives only in their head, and the patterns they have learned from cases that went wrong.
Constraints. It must be searchable, it must be maintainable in under fifteen minutes a week, and it must contain no identifiable information about any person. That last constraint is not negotiable and should be stated firmly, since the exercise otherwise invites a serious data protection breach.
Recording systems in health and social care were largely built to defend organisations rather than to support thinking, and the evidence for that is in their design. They are optimised for retrieval by administrators, for compliance reporting and for demonstrating that a required action occurred. They are not optimised for a practitioner trying to work out what is going on. Munro's review of child protection made the argument that increasing proceduralisation had displaced professional judgement rather than supporting it, and every practitioner in the room will have a version of the same complaint.
Concede the point fully, then press on the part the group controls. The chronology and the personal index are both things a practitioner can build without permission, and both are genuinely additional work in a job that has no spare hours. Ask the group what they would stop doing to make room. If the honest answer is nothing, the module has produced a good intention and not a change, and it is better to know that.
Build a full multi-source chronology on one open case, to the standard set today. Submit with a five hundred word analysis of what it showed. This is a required portfolio item.
A Scandal in Bohemia, the docketing passage. Munro, The Munro Review of Child Protection, final report, 2011, the chapters on recording and proceduralisation. Vannevar Bush, "As We May Think", 1945.
Thinking aloud, structured challenge and the practitioner's own feelings
"I am lost without my Boswell." A Scandal in Bohemia
To establish articulation and challenge as working procedures rather than as good manners, and to treat the practitioner's affective response as evidence to be examined rather than noise to be suppressed.
Aviation solved a version of this problem and left a written record of how. The accident literature of the nineteen seventies contains a series of crashes in which a junior officer knew and did not say, or said it too gently to be heard, and the response was crew resource management and the two-challenge rule, under which a crew member who has raised a concern twice without adequate response is required to take over. General practice arrived somewhere similar by a different route with Balint groups, which since the nineteen fifties have brought doctors together specifically to examine their own emotional reactions to patients on the premise that those reactions are clinical information. Software engineers rediscovered the whole thing as rubber-duck debugging, which is the observation that explaining a problem aloud to an inanimate object solves it, because articulation is where the flaw becomes audible. Four professions, no contact, one finding. Reasoning that stays inside one head stays wrong.
| Time | Activity |
|---|---|
| 0.00 | Chronologies returned. Two presented in full to the group. |
| 0.25 | Input. Articulation, challenge, the hierarchy gradient. |
| 0.45 | Exercise 9.1 Four minutes to a stranger. |
| 1.30 | Break. |
| 1.50 | Exercise 9.2 The challenge protocol. |
| 2.35 | Exercise 9.3 What do I feel about this case. |
| 3.15 | Counterpoint. When two heads are worse than one. Close. |
Time 45 minutes. Grouping Pairs, cross-profession, rotating twice.
Each learner presents a live case from their own work in four minutes. The listener is from a different profession and is permitted exactly three interventions, all of which must be one of the following.
Nothing else. No advice, no sympathy, no comparable case of their own. Three rotations of fifteen minutes.
Facilitator note. The constraint on the listener is what makes this work, and listeners will break it constantly in the first round, because offering help is what these professions are built to do. Stop the room after round one and reset. By round three most pairs will have found that the second question is the one that lands, since practitioners very rapidly discover how much of what they know about a case they were told by someone who was told it by someone else.
Debrief question. How far back does the evidence for your central claim actually go, and did you ever check.
Time 45 minutes. Grouping Fours, mixed profession and mixed seniority.
Introduce a two-part form and drill it until it is automatic. The form matters because challenge fails on delivery far more often than on content.
Advocacy. "I am seeing X. Here is the evidence I am reading it from."
Inquiry. "How are you seeing it. What am I missing."
Escalation, where the concern is not resolved. "I have raised this twice. I remain concerned about X for reason Y. I want it recorded that I have raised it, and I want us to agree what would settle it."
Each learner brings the case from their Module 7 alternatives worksheet. The group challenges it in turn using the form. Ten minutes each.
Two rules. The person challenged may not respond for the first two minutes. And the challenge must be delivered to the reasoning rather than to the decision, which means the phrase "I would have done X" is barred.
Facilitator note. Mix seniority deliberately and then debrief the effect of it openly. Ask who found it harder to challenge and why. The hierarchy gradient in health and social care runs along professional lines as well as along grade lines, and the group will name pairings where challenge is effectively impossible. Write them on the board. Those pairings are a live risk in every one of their organisations and naming them is more useful than pretending otherwise.
Time 40 minutes. Grouping Fixed reflective groups of five, which continue for the rest of the programme.
Run on the Balint pattern, adapted. One learner presents a case they find difficult, briefly and without notes, and then withdraws from the conversation and listens while the group discusses not the case but their own responses to it. The presenter rejoins at the end.
Prompts for the discussion. What did I feel as I listened. Who did I find myself siding with. Which person in this account did I stop thinking about. What would I not want to be the one to do here.
Framing input, five minutes before starting. Holmes is explicit that the emotional qualities are antagonistic to clear reasoning, and on this the canon is straightforwardly wrong for our purposes. The practitioner's reluctance to visit, the disproportionate liking for one parent, the irritation with a patient who is not obviously irritating, all of these are among the earliest signals available and all of them are routinely discarded as unprofessional. The discipline is not to act on the feeling. It is to notice it, name it, and ask what it is responding to.
Facilitator note. These groups must be fixed for the remainder of the course and must not include line managers of any member. Confidentiality terms should be agreed in writing at the start.
Groups do not reliably improve reasoning and sometimes degrade it. They converge, they suppress the dissenting member, they generate confidence faster than accuracy, and a supervision relationship of long standing very easily becomes a machine for confirming a shared view of a family. Multi-agency meetings are particularly prone to this, since the decision has often been made in a corridor beforehand and the meeting exists to record it.
There is also a specific objection to today's canonical model. Watson very seldom changes Holmes's mind, and it is hard to find a case in the canon where an objection of his alters a conclusion. He asks questions that permit Holmes to explain himself, which flatters the reasoning rather than testing it, and a supervisor who performs the Watson function in that sense is worse than useless because the practitioner leaves feeling examined. Ask the group to distinguish, from their own experience, supervision that tested them from supervision that let them rehearse. Most people can name both and the difference is usually a single question.
Deliver two challenges in your own workplace using the form, and receive one. Log all three, including what changed. Minimum six entries required by week 26.
A Scandal in Bohemia. An account of crew resource management and the two-challenge rule. Any short introduction to Balint groups. Ferguson on the emotional and embodied dimensions of home visiting.
Weighing, calibrating confidence and crossing a threshold
"Circumstantial evidence is a very tricky thing. It may seem to point very straight to one thing, but if you shift your own point of view a little, you may find it pointing in an equally uncompromising manner to something entirely different." The Boscombe Valley Mystery
To give the cohort a shared language for confidence, and to work through the question of when a belief obliges an action.
Sherman Kent ran analysis at the CIA in the nineteen fifties and became exercised by a report that said an attack was a serious possibility. He asked his colleagues what number they had in mind. The answers ranged from twenty per cent to eighty. He then published a scale of words with numbers attached, and intelligence services have used versions of it ever since, because the alternative is a system in which the analyst and the reader use the same word to mean opposite things and neither discovers it. Weather forecasting went further and became the best calibrated profession there is, for one reason. Forecasters find out the next day whether they were right, every day, for a career. Set that beside Meehl's finding in 1954, replicated many times since, that simple actuarial formulae outperform expert clinical judgement across a wide range of prediction tasks. The lone expert reading the signs is the most attractive figure in this course and the evidence about him is not kind.
| Time | Activity |
|---|---|
| 0.00 | Challenge logs. Plenary on what was hard. |
| 0.20 | Exercise 10.1 What does "likely" mean. |
| 0.55 | Input. Kent's scale. Standards of proof. Meehl. |
| 1.20 | Break. |
| 1.40 | Exercise 10.2 The cost matrix. |
| 2.25 | Exercise 10.3 The threshold staircase. |
| 3.10 | Counterpoint. Holmes decides alone. Close. |
Time 35 minutes. Grouping Individual then plenary.
Give the group fifteen phrases drawn from their own recent case records and reports. Each learner writes the percentage probability they take the phrase to mean. Collect and plot.
Phrases to use: likely, unlikely, a significant risk, cannot be ruled out, some concerns, a real possibility, probable, low risk, may, there is a risk that, no immediate concerns, appears to be, suggestive of, consistent with, we remain concerned.
The result. Ranges are enormous and overlapping. In most cohorts "a real possibility" spans fifteen to seventy five per cent, and "cannot be ruled out" ranges from one per cent to fifty. Two phrases will usually invert, so that some learners read phrase A as more probable than phrase B while others read the reverse.
Then build the scale. The group agrees a local lexicon of six or seven terms with numerical bands attached, writes it on a card, and undertakes to use it. Ten minutes. The agreement matters more than the exact numbers.
Facilitator note. Point out that "consistent with" is the most dangerous phrase in the set, because it is true of almost anything and reads to a non-specialist as confirmation. It appears constantly in medical reports and in court, and it is doing far more persuasive work than its logical content warrants.
Time 45 minutes. Grouping Mixed fours.
For a given decision, groups complete a four-cell matrix and populate every cell with specific consequences to specific people rather than with abstractions.
| The concern is real | The concern is not real | |
|---|---|---|
| I act | Harm prevented. What was the cost of acting anyway | Harm caused by acting. To whom. For how long |
| I do not act | Harm not prevented. To whom | Correctly left alone. What did that preserve |
Facilitator note. The top right cell is the one professions systematically under-populate. The costs of intervening wrongly are diffuse, delayed and borne by people who rarely complain in a form that reaches the organisation, whereas the costs of failing to intervene arrive in a review with the practitioner's name in it. That asymmetry is not a personal failing, it is a structural incentive, and naming it as such is more useful than urging people to be balanced.
Second round. Redo the matrix from the position of the person concerned rather than the professional. The cells often reverse in weight, and that reversal is the beginning of the argument that Module 12 completes.
Time 45 minutes. Grouping Threes then plenary.
Procedure. A scenario is released in nine steps, each adding a small increment of concern. After each step every learner privately marks the action they would now take from a fixed list. Take no action. Record only. Discuss in supervision. Make an unannounced visit. Consult another agency. Formal referral. Immediate protective action.
Plot the room. The spread is always wide and it is always wider between professions than within them.
Debrief questions.
Holmes releases a thief in The Blue Carbuncle on the grounds that prison would ruin him. He conceals a killing in The Abbey Grange and appoints himself and Watson as judge and jury in so many words. He commits burglary in Charles Augustus Milverton and watches a murder without intervening. In each case his judgement is presented as superior to the law's, and in each case the reader is invited to agree.
No practitioner in the room has that authority and every one of them will feel the pull of it, usually when a threshold is about to produce an outcome they think is wrong. That is the moment worth examining. Ask the group for an occasion when they believed the correct professional decision was the wrong human one, and what they did. The answers are usually careful and they are frequently the most valuable thing said in the whole programme.
Write one threshold note on a live decision, using the agreed confidence lexicon and a completed cost matrix. Bring to Module 11.
The Boscombe Valley Mystery. The Blue Carbuncle, The Abbey Grange and Charles Augustus Milverton, for the counterpoint. Sherman Kent, "Words of estimative probability", 1964. Meehl, Clinical Versus Statistical Prediction, 1954, or a summary of the replication literature.
Communicating reasoning in writing, in meetings and in court
"I had," said he, "come to an entirely erroneous conclusion which shows, my dear Watson, how dangerous it always is to reason from insufficient data." The Speckled Band
To produce written and spoken accounts in which a third party can audit every step from evidence to conclusion.
The methods section of a scientific paper exists so that a stranger can repeat the work and get the same answer, and the same convention governs an air accident report, a judgment and a well-written mathematical proof. Richard Feynman appended a personal statement to the Challenger report because the main text had smoothed the reasoning into consensus, and his appendix is worth circulating for its last sentence about nature not being fooled. All of these are documents whose authority comes from being checkable rather than from being confident. Case records almost never work this way. They record what was concluded and very rarely how, which means the next reader must either accept the conclusion or start again, and under pressure they accept it.
| Time | Activity |
|---|---|
| 0.00 | Threshold notes. Three read aloud and examined. |
| 0.25 | Input. Auditable writing. The four defects. |
| 0.45 | Exercise 11.1 The audit test. |
| 1.25 | Break. |
| 1.45 | Exercise 11.2 Rewriting to show the working. |
| 2.25 | Exercise 11.3 The simulated meeting. |
| 3.10 | Exercise 11.4 Would you say it to their face. Close. |
Time 40 minutes. Grouping Pairs, cross-profession.
Each learner hands over a real anonymised assessment they wrote. The partner performs an audit with a highlighter and four colours.
Count the four categories and calculate the proportion in the second and fourth. Write the ratio on the board anonymously.
Facilitator note. The ratios are usually poor and the room usually goes quiet. Resist reassurance. Then make the point that matters, which is that these documents are the basis on which other professionals act, and that an unsourced claim in a case file acquires the status of fact by the third time it is copied forward. Every practitioner in the room has both written one of these and acted on one.
Time 40 minutes. Grouping Individual then pairs.
Learners rewrite one section of their audited assessment to the programme standard. Teach the standard as four moves, and give a worked example of each.
| Move | Instead of | Write |
|---|---|---|
| Source the claim | "The home was cold." | "The living room was cold when I visited at 2pm on 14 May. The heating was off. Mrs A said she turns it on at four." |
| Separate report from event | "Mother had been drinking." | "The neighbour told the police she believed Mrs A had been drinking. I did not observe this and have not spoken to the neighbour." |
| Show the inference | "There are concerns about neglect." | "I saw X and Y. These would be consistent with Z. They would also be consistent with W. I currently think Z more likely because of V." |
| State confidence and the defeater | "It is likely that..." | "On the agreed scale this is probable, around seventy per cent. I would revise this if the school reports normal attendance this term." |
Objection to anticipate. Somebody will say there is no time. The answer is that the third move usually shortens the document, because the sentence that names its evidence tends to replace three sentences of atmosphere. Have a worked before and after ready to prove it.
Time 45 minutes. Grouping Whole group in role.
A multi-agency meeting is convened on the Module 8 case. Roles are allocated in advance, including a chair, representatives of five agencies, a family member and one participant briefed to hold a strong and wrong position confidently. Learners must present their reasoning and respond to challenge in role.
Observation focus. Two learners observe rather than participate, recording three things. Where did an unsourced claim enter the discussion and did anyone ask. Who spoke least and whose account was not sought. At what minute did the meeting settle on its conclusion, measured against the minute it formally reached it.
Debrief. The observers report first and their findings are usually more useful than the participants' reflections. The gap between the minute the meeting settled and the minute it concluded is normally substantial, and everything in between was ratification.
Time 20 minutes. Grouping Individual, silent.
Each learner reads their rewritten section imagining the person it concerns is reading it with them. They mark anything they would not say aloud to that person and rewrite it, without softening the substance.
The distinction to hold. This is not an instruction to be kind at the expense of accuracy. It is a test of whether a judgement can survive being stated openly to the person it is about. A judgement that cannot survive that is usually one that rests on something the writer has not examined, and the discomfort is diagnostic rather than merely social.
The objection is real and practitioners raise it for good reasons. A document that displays each inferential step gives a solicitor twelve places to attack rather than one, and a practitioner who records a considered alternative that later proved correct has written the sentence that will be read aloud at their hearing. Defensive vagueness is not stupidity. It is a rational response to an environment that punishes visible reasoning more reliably than it punishes bad reasoning.
Say this plainly, then put the counter. Vagueness protects the individual in the short term and it degrades every subsequent decision made on the record, including decisions about the same person years later by people who never met the writer. There is no version of this where the practitioner alone can fix it. Ask the group what their organisation would have to do differently to make transparent reasoning safe, and send the answers to whoever commissioned the course.
Rewrite one complete assessment to the standard. Submit both versions with a short commentary. Complete the portfolio for submission at week 26.
The Speckled Band. Feynman's personal appendix to the Rogers Commission report on the Challenger accident, 1986. One judgment or air accident report, read for how it shows its working.
Where the method fails, and where it harms
"Watson," said he, "if it should ever strike you that I am getting a little over-confident in my powers, or giving less pains to a case than it deserves, kindly whisper 'Norbury' in my ear, and I shall be infinitely obliged to you." The Yellow Face
To dismantle, with the cohort's participation, the model they have spent five months learning, and to establish what survives the dismantling.
This module is the reason the programme can be taught at all. A course that produced practitioners more confident in their own perception would increase harm, and a good deal of the harm done by health and social care is done by people who were sure they could read a situation. Deliver this module with the same seriousness as the rest and give it the second facilitator and the service user co-facilitator it requires.
Doyle set the clue after he knew the answer, so the evidence in a Holmes story has exactly one plausible parent, and real evidence never does. Any set of present traces is compatible with an indefinite number of past histories, and the method has no way of telling you when you are in a case that will not resolve. Holmes is never given an unsolvable one.
The reflexive version is more uncomfortable and the group should sit with it. Serious case reviews are also written backwards, by people who know how it ended, and they routinely produce a narrative in which the pattern was legible all along. That narrative is then used to train the next generation of practitioners in what to look for, which means the profession is partly being trained on retrospectively manufactured coherence. Hindsight is the most reliable producer of Holmesian clarity there is, and it has never yet predicted anything.
A tanned man with a stiff arm has come from Afghanistan. He might equally have come from anywhere hot and been injured anywhere at all, and the inference is only compelling because the story has been arranged around it. Nowhere in sixty stories does Holmes ask how common something is, which is the first question any sound inference requires. Module 2 established that the practitioner's sense of frequency comes from a caseload rather than from a population. That defect is invisible from the inside and it is the mechanism by which experienced practitioners become confidently wrong about rare conditions and about common ones.
This is the objection with the most evidence behind it and the one the cohort will find hardest, because the whole appeal of the material runs the other way. Meehl's review in 1954 found that simple actuarial rules matched or beat expert clinical judgement, and the finding has survived six decades of attempts to overturn it across a wide range of prediction problems. Structured professional judgement instruments in risk assessment exist for this reason. Checklists in surgery exist for this reason. The debiasing literature, such as it is, finds its most consistent effects in guided procedures rather than in cultivated insight.
What follows is not that judgement is worthless. It is that judgement performs best when it is doing what instruments cannot, which is deciding what information to gather, noticing the case that does not fit the instrument, and interpreting the instrument's output for a particular person. Holmes's actual advantage over Lestrade is not that he is cleverer. It is that he goes to the house, looks at everything, and considers more than one explanation. Every one of those is proceduralisable, and it is precisely the proceduralisable part that this course has been teaching.
Holmes's method is a direct descendant of Victorian physiognomy, and the canon carries the period's prejudices without embarrassment. Tonga in The Sign of Four is described in terms that are straightforwardly racist. The Adventure of the Three Gables opens with an exchange that is worse. Class inference runs through everything, and it is presented as observation.
The professional danger is exact rather than general. A practitioner trained to read circumstance from appearance, and rewarded socially for doing it well, is a practitioner equipped to produce discriminatory conclusions in the grammar of evidence. Disproportionality in child protection and in mental health detention is not produced by practitioners who believe themselves to be prejudiced. It is produced by practitioners exercising professional judgement on cues that correlate with race and class, and writing it up as assessment.
The safeguard is not an attitude. It is the three-column discipline from Module 3 and the alternatives requirement from Module 7, applied hardest at exactly the moment when a reading feels most obvious.
Holmes infers about people and almost never with them. His clients supply information and receive conclusions. The relationship is one of interpretation, and the authority over the meaning of a life sits wholly with the interpreter.
Contemporary practice is built on the opposite claim. The person has knowledge of their own situation that the professional does not and cannot have, and that knowledge has standing rather than merely being data. The disability rights movement, the survivor movement in mental health and the co-production literature have all made this argument from different directions and it is not a soft one. It is an epistemic claim, namely that professional inference about a life is systematically worse than the account of the person living it, in most respects, most of the time.
Both claims cannot be fully right. Hold the tension rather than dissolving it. The programme's position, which the group is free to reject, is that Holmesian method has a proper place where the person cannot speak, will not speak, or is not the only person concerned, and that it becomes an intrusion the moment it is used on someone who is telling you the answer.
Already covered in Module 9 and worth restating here. Holmes's manifesto is wrong for these professions, and his practice contradicts his manifesto more often than he admits.
This objection has no literature behind it and it may be the most important one in the room. The material is enjoyable. Being right about a stranger from small signs is a genuine pleasure and it is the pleasure that sells the course. It is also a corrupting one, because the practitioner who enjoys reading a situation will read situations that are not there, will prefer the elegant explanation to the dull one, and will experience being challenged as a loss. Every one of those is a documented failure pattern in the reviews.
The undertaking made in Module 0 comes back here. When the room enjoys itself, say so.
| Time | Activity |
|---|---|
| 0.00 | Read the Norbury passage. Facilitator input on the seven objections, twenty five minutes, no discussion yet. |
| 0.25 | Exercise 12.1 The prosecution of Sherlock Holmes. |
| 1.20 | Break. |
| 1.40 | Exercise 12.2 Reading the canon's prejudice. |
| 2.15 | Exercise 12.3 The other chair. Service user and carer panel. |
| 3.00 | Exercise 12.4 What survives. Taking down the Module 0 sheet. |
| 3.25 | Capstone briefing. Close. |
Time 55 minutes. Grouping Two teams plus a bench of three.
A formal hearing. The charge is that the method taught on this programme is unsafe for use in health and social care. Prosecution and defence each have fifteen minutes to prepare, ten minutes to present and five to reply. The bench of three questions both and delivers a verdict with reasons.
Rules. Each side must use at least three of the seven objections and must cite specific canon or specific evidence. Assertion without a source is disallowed by the bench. The defence may not argue that the method is merely a bit of fun, and the prosecution may not argue that professional judgement should be abolished. Both of those are available and both are lazy.
Facilitator note. Allocate the most enthusiastic learners to the prosecution. The ones who have most enjoyed the programme produce the best case against it, and arguing it changes them in a way that hearing it does not. The bench should include the service user co-facilitator.
Time 35 minutes. Grouping Threes.
Distribute four short passages in which Holmes or Watson infers character or worth from appearance, race or class. Groups do two things with each.
Facilitator note. The second task is the exercise and groups find it easy, which is the point. "Presented as unkempt." "Home conditions were chaotic." "Mother was hostile." "Father was difficult to engage." "The family are known to services." Every one of these is a physiognomic judgement in professional clothing, and every one of them has appeared in a file that ended badly for someone.
Closing question. Which of these have you written. Silence is fine as an answer.
Time 45 minutes. Grouping Whole group with a panel of three.
A panel of people with lived experience of being assessed. Brief them in advance on what the cohort has been taught, in detail, and pay them properly. The session is theirs rather than the facilitator's.
Questions to seed, if seeding is needed. What did you know that nobody asked. What did a professional conclude about you from something they saw, and were they right. What did it feel like to know you were being read. What would you have wanted them to do instead.
Facilitator conduct. Do not defend the profession and do not defend the course. If a panel member describes something the cohort has been taught to do and describes it as intrusive, let that stand unanswered. The urge to explain is strong and giving in to it wastes the session.
Time 25 minutes. Grouping Whole group.
Take down the sheet from Module 0 with the question about Holmes carrying a caseload of thirty two. Read it out. Then the group builds two lists on facing walls.
Facilitator note. The keep list is almost always procedural. The three-column record, the chronology, the alternatives requirement, the four minutes of free recall, the confidence lexicon, the trigger to revise, the challenge form. The discard list is almost always the glamorous material, and specifically the trace-reading virtuosity from Module 3. That result should not be smoothed over. It is the programme's honest finding about itself, and a cohort that reaches it has learned the thing the course exists to teach.
Complete the 1,500 word critical reflection. Submit the full portfolio by week 26.
The Yellow Face. The Sign of Four and The Adventure of the Three Gables, for the canon's prejudice. Meehl, and one recent paper on disproportionality in the learner's own sector. Anka and colleagues, Professional Curiosity in Safeguarding Adults, 2025.
Assessed simulation
A full day of six hours rather than a half day. Learners are assessed individually across four stations of forty minutes, with rotation, briefing and two breaks accounting for the remainder. The written product is produced within stations one and four rather than at a station of its own. Two assessors per station, one of whom must be from a profession other than the candidate's, and one station is co-assessed by a person with lived experience.
| Station | Task | Competencies | Weight |
|---|---|---|---|
| 1. The visit | Twenty minutes in a dressed set with an actor. Structured observation and initial elicitation. Followed by twenty minutes writing the three-column record. | A1, A4, C1, C2, C3 | 25% |
| 2. The bundle | A forty page multi-agency bundle. Produce a chronology and an absence audit in forty minutes. | A3, D1, D2 | 25% |
| 3. The reasoning | Complete a reasoning trace on the material from stations one and two. Defend it under challenge from both assessors for fifteen minutes. | B1 to B6, D3, E1 | 25% |
| 4. The decision | State the action you would take, the threshold applied and the confidence held. Then write the record entry a colleague would rely on. | E1, E2, E3 | 25% |
The scenario must have the following properties, and building it takes about three days of facilitator time. Reuse it across cohorts with variations.
| Criterion | Fail | Pass | Distinction |
|---|---|---|---|
| Observation | Records conclusions as observations. Misses most planted detail | Uses a protocol. Separates observed from inferred consistently | Recovers subtle detail and at least two absences. Notes own affective response |
| Inference | Single explanation, held as fact | Two or more explanations, correctly labelled as abduction, with discriminating information identified | Ranks explanations against stated evidence and identifies what was never on the list |
| Elicitation | Interrupts free recall. Leading or contaminating questions | Completes free recall, probes without leading, summarises back | Adapts register appropriately. Correctly classifies non-disclosure and responds to the class it is in |
| Organisation | No usable chronology. Sources not traceable | Chronology to standard, sources traceable, report distinguished from event | Derives an analytic finding from the chronology that the narrative did not contain |
| Judgement | Confidence unstated or miscalibrated. Threshold not articulated | Uses the confidence lexicon. States threshold and standard of proof | Populates the full cost matrix including the costs of acting wrongly |
| Communication | Unsourced claims. Conclusions without working | Every claim sourced, every inference evidenced | Record is legible to a stranger and could be read by the person it concerns |
| Discrimination | Builds on a class or race-coded cue. Reads poverty as neglect | Identifies the cue and declines to build on it | Names the operation explicitly in the record and states the safeguard applied |
A fail on the discrimination criterion is a fail overall, irrespective of other marks. This should be stated in Module 0 and repeated in Module 12.
A generic template. Learners build their own in Module 1 and this is the fallback for those who cannot. Print at card size. The order is fixed and should be run the same way every time, because a variable order defeats the purpose.
Items 11, 12 and 13 are mandatory in every local variant. Everything above them may be replaced.
Replace items 1 to 10 with: how the patient entered and sat, hands, breathing at rest, what they brought with them, who came with them and who spoke, what they said first before you asked, the gap between what they came for and what they have talked about, medication actually present, what they have not mentioned that this condition usually brings, and the point at which you formed your working diagnosis and what time that was.
Replace items 1 to 10 with: the bedside and what is on it, visitors' traces, the call bell and whether it is within reach, the fluid chart against the fluid, personal belongings, what the person is wearing and who chose it, how staff refer to them when they are present, how staff refer to them when they are not, the last three entries in the notes and who wrote them, and who has not visited.
| Date of event | Date recorded | Source | Event, neutral language | Present / absent | Status | Significance |
|---|---|---|---|---|---|---|
| 14/05/24 | 02/06/24 | HV record, p.12 | Home visit. Front door not answered | HV. No adult seen | Fact | Third of four |
Rules. One line per event. Status is fact, report or inference and nothing else. A report of an event is recorded as a report and the reporter is named. The significance column is completed on a second pass only. Do not merge sources. Do not remove a line because it seems unimportant, since importance is what the chronology is for and you cannot know it while building.
Agreed locally in Module 10. The bands below are a starting point drawn from the intelligence analysis convention and should be amended by the cohort rather than imposed.
| Term | Band | Use when |
|---|---|---|
| Remote | under 5% | Considered and effectively excluded, but not impossible |
| Unlikely | 10 to 20% | Evidence points away but the possibility is live |
| Realistic possibility | 25 to 50% | Cannot be discounted and would change the plan if true |
| Likely / probable | 55 to 75% | The best current explanation with competitors still standing |
| Highly likely | 80 to 90% | Competitors considered and each has specific evidence against it |
| Almost certain | over 95% | Direct evidence, corroborated, no live alternative |
The gaps between the bands are deliberate and the group should be told so when the lexicon is agreed, since a cohort will otherwise read them as an arithmetic slip. Five short stretches are left empty, each of them five points wide. A scale with no dead ground lets a practitioner park on a boundary and call the parking a judgement. Where an estimate falls in a gap, the term has not yet been chosen and the evidence has to be looked at again. Note which side of realistic possibility the dead ground falls on. It sits below that band rather than above it, because realistic possibility is the term that changes the plan, and ground taken off its floor would make a case easier to dismiss.
Barred without qualification: consistent with, used alone. It is true of almost everything and reads as confirmation. If used, state what else it would also be consistent with, in the same sentence.
| Item | Length | Due | Pass standard |
|---|---|---|---|
| 1. Baseline and closing self-audit with commentary | 800 words | Wk 26 | Specific change identified and evidenced, not asserted |
| 2. Three observation records, three-column | - | Wks 5, 9, 21 | Columns properly separated. One must record an inference that proved wrong |
| 3. Multi-source chronology with analysis | 500 words plus chronology | Wk 19 | Status column correct throughout. Analysis derives something the narrative did not contain |
| 4. Reasoning trace on a live decision | - | Wk 21 | Field six written before the outcome was known, with evidence of the date |
| 5. Peer challenge log | Six entries minimum | Wk 26 | Both given and received. Each entry states what changed, including "nothing" |
| 6. Critical reflection on the limits of the method | 1,500 words | Wk 26 | Engages at least three of the seven objections with reference to the learner's own practice |
Item 2's requirement that one record show a wrong inference is the hardest to enforce and the most important. Learners will submit three successes. Return them. A programme that cannot produce evidence of learners being wrong has not taught anything about being wrong.
For services that cannot release staff for the full programme. Each runs in under an hour in a team meeting, and the first six work as a standalone six-week series.
Modules 1, 5 and 12 compressed. Attention, absence and the critique. Nothing else survives compression and attempting more produces a day of entertainment. Use exercises 1.1, 1.3, 5.1, 5.3, 12.2 and 12.4.
Day one, Modules 1 and 2. Day two, Modules 5 and 7. Day three, Modules 9 and 12. Workplace tasks between. Assessment by portfolio items 2, 5 and 6 only. This version keeps the procedural content and drops the trace-reading, which is the right sacrifice, since the trace-reading is the most enjoyable material and the least defensible.
The programme requires build time that commissioners routinely underestimate. Budget for it.
Conan Doyle's letter to Joseph Bell of 4 May 1892, and the Royal College of Physicians of Edinburgh's material on Bell, for the origin story used in Module 0.
The programme will be judged by commissioners on whether learners enjoyed it, and they will. That is not the measure. The measure is whether, six months after the capstone, anyone is still building chronologies, still writing an alternative next to every inference, and still able to say out loud that they do not know.
Holmes gave Watson a word to whisper when he was getting a little over-confident in his powers. He did it after the one case in the canon he got badly wrong, and he did it because he understood that the corrective would have to come from outside him and would have to be arranged in advance. That is the most useful thing in sixty stories, and it is the only piece of the method that costs nothing to adopt.