Hypothesis-Tree Review¶
Structured review — instantiates Constraint-Guided Backtracking
A structured human checkpoint that walks the tree of live and refuted hypotheses, judges which branches are genuinely closed, and chooses where to resume or when to escalate.
A Hypothesis-Tree Review is a human checkpoint over a reversible investigation: a structured, periodic review in which people — not an algorithm — walk the current tree of hypotheses, judge which branches the evidence has genuinely refuted versus merely made unattractive, decide the point to resume from, and determine when the search should be handed to a person with authority or to a different method. Its defining idea is that human judgment governs the search precisely at the moments a mechanical failure test cannot be trusted — where evidence is ambiguous, stakes are high, or a rule at issue is a real-world constraint rather than a mere search preference. It is not a faster test; it is the deliberate insertion of a person where automation would over-commit.
Example¶
After a turbine trips on a power plant, an engineering team runs a root-cause review over a fault tree of candidate causes: sensor fault, control-loop instability, mechanical imbalance, foreign-object damage. Several branches were pursued and quietly set aside during the first day of investigation, and the review reopens each one for scrutiny.
"We ruled out imbalance on a single vibration reading — was that a genuine refutation or just a weak signal on a noisy channel?" The team decides imbalance was only doubted, not disproven, and reopens the branch. They agree the control-loop branch is genuinely dead, because a bench test reproduced its behavior and cleared it. They then choose the resume point — not the most recent thing anyone tried, but the last fork everyone agrees is solid, which sits several steps back. And because a safety-relevant cause cannot be dismissed on one test, they escalate the sensor branch to an independent inspector rather than let the investigation close it on its own authority. The review made no measurements; it adjudicated which closures to trust, where to restart, and what to lift out of the team's hands entirely.
How it works¶
- Convene at a checkpoint. On a cadence or a trigger, pause the search and bring the current hypothesis tree in front of reviewers.
- Adjudicate each failure claim. For every branch marked dead, ask whether the evidence is a genuine refutation or merely a weak or convenient reason to abandon it; reopen the ones that fail the bar.
- Choose the resume fork. Decide where to restart — often deeper than the latest choice, at the earlier decision actually implicated, rather than the most recent one.
- Escalate or hand off. When judgment, authority, or safety requires it, route a branch to a person or a different method instead of letting the automated search close it.
Tuning parameters¶
- Cadence — review on a fixed interval or only on a trigger (a surprising result, a stalled search). Frequent reviews catch bad closures early but cost attention.
- Authority / quorum — who is allowed to declare a branch genuinely closed, and how many must agree. Higher bars resist premature closure but slow the search.
- Evidence bar — how strong a refutation must be to count as genuine. A high bar reopens more branches; a low one converges faster but risks dismissing a live cause.
- Escalation threshold — what conditions force a branch out of the automated search and into human or independent hands.
- Reopen policy — how readily set-aside branches are revisited, trading thoroughness against ever converging.
When it helps, and when it misleads¶
Its strength is rescuing a search from mechanical over-confidence and from unsafe auto-closure: it catches the branch that was abandoned on a weak signal, and it stops an algorithm from closing a safety- or rights-bearing question it has no business closing.
Its defining failure mode is bias in the adjudication — reviewers close the branch they want closed and keep alive the one they favor, so the review launders a hunch as a verdict; the standard corrective is a structured "consider-the-opposite" discipline and a designated challenger.[n1] The mirror-image misuse is review theater (rubber-stamping whatever the search already decided) or its opposite, endless reopening that never lets the investigation converge. The guarding discipline is to require an explicit evidentiary reason to close any branch and to assign someone to argue the other side, so the review adds judgment rather than merely ceremony.
How it implements the components¶
The Hypothesis-Tree Review fills the human-judgment slice of the archetype — the checkpoint where a person overrides or ratifies the mechanical search:
human_review_escape_hatch— the review is the escape hatch: the defined point at which the search is lifted into human hands or handed to an independent authority.constraint_failure_test— reviewers apply the failure test as human adjudication, distinguishing a genuine refutation from mere doubt, dislike, or a weak signal — the judgment a mechanical check cannot make.rollback_rule— the review sets the resume point, and unlike chronological return it can justify a deeper step-back to the earlier fork actually responsible for the trouble.
It does not run a mechanical look-ahead over undecided parts — forward_checking_probe belongs to Forward-Checking Table, its nearest twin among the archetype's tests; that table is an automated, predictive wipeout check, whereas this review is a human, retrospective judgment about genuine refutation. Nor does it guarantee the resumed state can be restored intact — preserved_context_boundary is Undo-Stack Protocol.
Related¶
- Instantiates: Constraint-Guided Backtracking — the review is the archetype's human checkpoint on an otherwise automated reversible search.
- Consumes: Decision-Tree Search Diagram — reviewers walk the drawn hypothesis tree to see the live and refuted branches they must adjudicate.
- Sibling mechanisms: Forward-Checking Table · Recursive Depth-First Backtracking · Undo-Stack Protocol · Chronological Backtracking Log · Decision-Tree Search Diagram · Constraint-Satisfaction Solver Pass
Editorial Notes¶
Form Classification¶
Form family: Decision, Gate & Allocation
Rationale: Hypothesis-Tree Review operates as a case-specific gate, selection, routing, prioritization, or resource disposition because it a structured human checkpoint that walks the tree of live and refuted hypotheses, judges which branches are genuinely closed, and chooses where to resume or when to escalate
Independent corroboration: The frozen evidence defines Hypothesis-Tree Review as 'A structured human checkpoint that walks the tree of live and refuted hypotheses, judges which branches are genuinely closed, and chooses where to resume or when to escalate', so its operative form is Decision, Gate & Allocation.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Security Studies & Intelligence Analysis
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Structured review of competing branches and disconfirming evidence descends from intelligence analysis of competing hypotheses.
Related originating lineages:
- Computer Science & Software Engineering — Tree search, backtracking, live versus closed branches, and resume points provide the computational structure.
- Organizational & Management Science — Scheduled expert review and escalation provide the institutional checkpoint.
- Psychology — Consider-the-opposite research materially supplies the confirmation-bias countermeasure.
- Statistics & Experimental Design — Model comparison and falsification provide an independent scientific review lineage.
Review resolution: CIA structured analytic tradecraft supplies the competing-hypothesis review lineage. Psychology contributes bias checks and tree-structured decomposition, making the encyclopedia review format a synthesis with security/intelligence primary. The retained alternate domains identify independent or materially shaping provenance, not downstream reach alone. domain_reach=multi_domain because the mechanism has independent established use in several fields. The encyclopedia entry deliberately composes those lineages.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
- https://www.cia.gov/resources/csi/static/Tradecraft-Primer-apr09.pdf — CIA tradecraft primer on structured analysis of competing hypotheses and bias.
Notes¶
[n1] Consider the opposite is a documented debiasing technique — deliberately asking why the favored conclusion might be wrong — used to counter confirmation bias, the tendency to weigh evidence toward a hypothesis one already prefers. Pairing it with a designated challenger is what keeps a hypothesis-tree review from ratifying the team's prior hunch. ↩