Hypothesis List¶
Template — instantiates Cautious Pattern Completion
Turns a single tempting explanation into an explicit slate of candidate completions drawn from the same partial input, so the first story cannot quietly become the only story.
A Hypothesis List is the plainest guard against single-story lock-in: from one bounded pool of partial evidence it writes down several candidate explanations at once, as a standing list, before any of them is investigated. Its single distinguishing move is plurality by default — at the moment of first encounter it treats the number of live explanations as something to be deliberately widened, not collapsed to the one that feels most coherent. It does not rank, weigh, or test; it does not decide anything. It only guarantees that the slate exists and stays visible, so that when investigation begins there is more than one story on the table to investigate.
Example¶
A growth analyst at a subscription app watches trial-to-paid conversion fall from roughly 22% to 15% over one quarter. The tempting single story arrives instantly: "the redesigned onboarding flow confused new users." Before pulling a single query, the analyst writes a Hypothesis List against the bounded input — the conversion drop plus the change log for exactly that quarter. Five candidates go on the list, each compatible with the same data: (a) the onboarding redesign, (b) a pricing-page experiment that shipped the same week, © a shift in ad spend toward a cheaper, lower-intent channel, (d) a billing integration that had begun silently declining certain cards, and (e) ordinary seasonality.
The list is posted in the incident doc before anyone starts digging. That ordering is the whole point: the billing bug (d) was nobody's favorite and would never have been raised in a conversation that opened with the onboarding theory. Because it sat on the slate as a peer, someone eventually checked it — and it was real. The Hypothesis List did not find the answer; it kept the answer from being ruled out before it was ever considered.
How it works¶
The template inverts the usual reflex, which is to converge. Its steps are deliberately incomplete on purpose:
- Bound the input first — write down exactly what is known and what window it covers, so every hypothesis is answerable to the same evidence rather than to a private hunch.
- Generate to a quota, not to comfort — commit to a minimum count (say, "at least four distinct explanations") so the list does not stop at the first plausible one.
- Keep candidates mutually distinct — prefer explanations that would be confirmed or refuted by different observations, so the slate spans the possibility space instead of restating one idea five ways.
- Withhold verdicts — the list carries no ranking, no confidence, no "leading candidate." It stops precisely where discrimination would begin, and hands off to a method that does that work.
Tuning parameters¶
- Breadth quota — the minimum number of live candidates. Raising it fights premature closure but eventually pads the list with implausibilities; lower it and the tempting story creeps back in as "the obvious one."
- Input-boundary tightness — how strictly each hypothesis must answer to the recorded evidence. Tight bounds keep the slate grounded; loose bounds invite free-form speculation.
- Mutual-exclusivity discipline — how hard you push for candidates that different evidence would separate. High discipline makes the later discriminating work cheap; low discipline yields near-duplicates that all survive every test.
- Generation mode — solo drafting versus an adversarial round where a second person must add a candidate the first would not. Adversarial generation widens the slate at the cost of time.
When it helps, and when it misleads¶
Its strength is entirely at first encounter: it is the cheapest possible interruption of the mind's rush to a single coherent story, and it is most valuable exactly when one explanation feels too obvious to bother listing alternatives — because that feeling is usually What-You-See-Is-All-There-Is at work, not evidence.[n1]
Its failure modes are two, and opposite. List inflation buries the two or three live explanations under a dozen trivial ones, so plurality becomes noise and the team defaults back to its favorite anyway. Decorative plurality is worse: the list is generated for form's sake, then everyone quietly works only the pre-favored candidate while the others sit unexamined as alibis for open-mindedness. The guarding discipline is to cap the slate at plausible-and-distinct, and to treat the list as an input to a discriminating method — never as a finished analysis. A list that is written and then never pruned by evidence has done half a job and claimed a whole one.
How it implements the components¶
partial_input— the first line of the template is the bounded statement of what is actually known and over what window, so every candidate answers to the same evidence.candidate_completion— each list entry is one possible coherent explanation, held explicitly as a possibility rather than asserted as the account.alternative_completion_set— the list as a whole is the maintained set of live alternatives; keeping the set plural is the mechanism's entire reason for existing.
It deliberately stops before discrimination: it does not name the missing_evidence_marker that would separate the candidates or update likelihoods as evidence lands — that is Diagnostic Differential, its nearest twin — nor does it hunt for a disconfirmation_probe against the leader (Disconfirming Evidence Search) or run the conclusion_hold_or_release_gate that decides release (Withhold-Conclusion Checkpoint).
Related¶
- Instantiates: Cautious Pattern Completion — the Hypothesis List supplies the plural slate the rest of the archetype's machinery narrows and gates.
- Sibling mechanisms: Diagnostic Differential · Disconfirming Evidence Search · Hallucination Check · Reconstruction Note · Source-Tracing Table · Uncertainty Tagging · Withhold-Conclusion Checkpoint · Assumption Log
Editorial Notes¶
Form Classification¶
Form family: Representation, Specification & Plan
Rationale: The template externalizes an explicit slate of rival completions from the same partial input so the first explanation cannot remain silently unique.
Nearest alternative: Interface, Display & Cue — The list cues broader thought, but its operative form is the durable representation of candidate hypotheses.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Security Studies & Intelligence Analysis
Origin pattern: Convergent development
Present-day reach: Universal
Rationale: Explicitly enumerating competing explanations before evaluating evidence is a core intelligence-analysis discipline.
Related originating lineages:
- Cognitive Science — Pattern completion and limited hypothesis search provide the information-processing account.
- Philosophy — Scientific and abductive reasoning independently require rival hypotheses rather than a single favored account.
- Psychology — Debiasing research supplies the WYSIATI and premature-closure rationale for forcing absent alternatives onto the page.
- Statistics & Experimental Design — Maintaining multiple candidate hypotheses is also foundational to disciplined empirical inquiry.
Review resolution: The CIA tradecraft primer explicitly requires analysts to generate and evaluate multiple hypotheses. Scientific and psychological inquiry independently use hypothesis lists, but this mechanism’s competing-explanation discipline directly follows intelligence analysis. The retained alternate domains identify independent or materially shaping provenance, not downstream reach alone. domain_reach=universal because the mechanism is portable across essentially any field. The entry generalizes an established mechanism without inventing a new cross-domain composite.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
- https://www.cia.gov/resources/csi/static/Tradecraft-Primer-apr09.pdf — CIA primary tradecraft guidance on generating and testing alternative hypotheses.
Notes¶
[n1] What You See Is All There Is — Daniel Kahneman's term for the mind's habit of building a confident, coherent story from only the evidence immediately in view, without registering what is absent. A Hypothesis List is a direct countermeasure: it forces the absent alternatives onto the page. ↩