The School of Unlearning — Facilitator's Field Guide

Version 2 / 23 October 2026 · Independent inquiry by Sahil Tayal. Eight original interactive studios, not a formal accredited curriculum.

What the school is for

The School of Unlearning was created around a simple educational disagreement: information transmission matters, but it does not exhaust the work of education. Learners also need opportunities to notice assumptions, ask whether an observation supports a claim, change a condition, compare explanations and decide what to retain. The school is an experiment in those intellectual habits, not an anti-knowledge manifesto. One can learn a factual principle with care and also learn to notice the limits of one's own application of it.

The rooms offer cases that cut across schooling, cognitive psychology, organisations, statistics, social interpretation, communication and human development. This interdisciplinarity is deliberate: ideas become durable when people can use them outside the scene in which they encountered them. But transfer should be tested, not asserted. The designer does not claim that using the site will improve standardised scores, cognitive ability or emotional wellbeing.

The five-part seminar (adapt to context)

• Predict without humiliation. Ask, “What do you expect, and what information are you using?” Invite “I do not know” as a legitimate starting point. Do not ask learners to reveal personal trauma, finances or relationships. Use the case as written or an invented alternative.
• Perturb one thing. Learners manipulate the actual studio instrument, select a source, or explore a decision branch. Ask them to say what they changed before describing the result.
• Distinguish observation from inference. “The ranking changed” is an observation. “The first student is more worthy” is an additional value judgment. “The detector's false-alarm proportion rose” is an arithmetic observation conditional on assumed parameters, not proof about a real detector.
• Steelman the objection. Ask a participant to make the best case for the position they currently oppose. A critique of grades should still acknowledge the public purposes of assessment; a critique of metrics should still acknowledge accountability.
• Transfer and disconfirm. “Where else would this matter?” is the transfer question. “What would make you change your mind once more?” is the disconfirmation question. There is no compulsory conversion to the project's preferred perspective.

Facilitation without grades

Do not give points for agreeing with the author. If an evaluation is mandatory, assess observable reasoning processes: whether the learner identifies a claim, grounds an observation, names alternative explanations, notices limits, and adapts a conclusion when warranted. Do not interpret reluctance to revise as intellectual incapacity; the learner may have a valid criticism of the model. Avoid performance leaderboards and public comparison of private notes.

An optional written reflection prompt: “At the beginning I expected __ because __. After changing __, I observed __. That supports __ but cannot establish __. I would now __. I would reconsider after seeing __.” One paragraph is sufficient; silence is permitted.

Studio-specific facilitation moves

01 · Scoreboard. Change selection criteria, keep the data stable. Make the strongest case for a precise shared evaluation standard. Ask what the rank omits, including opportunity and baseline measurement error. Do not turn the toy ranking into a system for grading real students.

02 · Evidence. Start by counting documents, then trace origins. Distinguish independence from quality. A tiny independent sample is not necessarily more useful than a substantial synthesis. Never conflate a repetition count with an effect size.

03 · Memory. Let people choose not to participate. One brief retrieval demonstration cannot estimate intelligence or retention gains. Contrast familiarity with recall, then design a genuine delayed-retention study on paper instead of claiming that the exercise proves the mechanism.

04 · Incentives. Make the invented equations visible. Invite critique of their coefficient choices. Ask what important outcome is missing. How could an organisation measure results with a balanced set of indicators, without building an impossible surveillance machine?

05 · Base rates. Use a group of 1,000 imagined items, not people or medical tests. Make denominators explicit. Explain why sensitivity and positive predictive value are different. Ask about the consequences of errors before recommending an operational threshold.

06 · Weather. Separate camera-recordable behaviour from feelings and explanations. Avoid asking learners to discuss actual conflicts. A boundary does not require certainty about another person's motives. The fictional branches never reveal a secret correct explanation.

07 · Persuasion. Hold evidence fixed while changing language. Ask which phrase converted an observation into a causal or population-wide claim. A concise caveat can be part of compelling communication.

08 · Garden. Require at least two explanations for a behaviour. Have learners choose an intervention and explain what observation would distinguish its proposed mechanism from alternative explanations. Do not reward punishment, support or personal responsibility as an ideology.

Short-format session (around one class period)

Opening question and private prediction; one room in small groups; a shared chart with columns for model input, observed output and limitation; two competing interpretations; individual transfer reflection. Time allocations depend on group and should not force an emotional disclosure. Longer workshops can pair Scoreboard with Incentives, Evidence with Persuasion, Memory with Base Rates, or Weather with Garden.

Accessibility and inclusion

Provide a written alternative to every spoken answer. The site supports keyboard-operated buttons, labels, an orientation dialog and reduced-motion preferences; nevertheless, facilitators should verify their own venue and assistive technology setup. The timed memory theatre may not suit everyone; learners may skip it, end exposure immediately or use the explanatory field sheet instead. Avoid assessing people by memory trial performance. Use the printable markdown lessons when technology is unavailable.

Privacy and safeguarding

The site does not call analytics or servers from its application code. The notebook persists locally in the browser when available; participants can export, import and erase entries. Shared devices create privacy risks: encourage users to close out or erase after export according to their own preferences. Never request personal stories for a public demonstration. Scenario participants and evidence are fictional composites, not identifiable acquaintances.

Evidence discipline

The design borrows broad principles from the learning-science literature, especially metacognitive monitoring, the utility of retrieval, and the role of context in learning. It is not itself validated as an intervention. In particular, no causal effect is implied by the Incentive Engine equations, the Scoreboard index or the Counterfactual Garden branches. See SOURCES_AND_LIMITS.md for exact provenance.

The most important closing question

Ask participants: “Which idea survived scrutiny—and why?” The purpose is not to maximise doubt. It is to make confidence more responsive to what can actually be known.