/formulate-hypotheses

name: formulate-hypotheses
description: "Generate testable hypotheses from theory, empirical regularity, anomaly, or explicit explanatory competition. Deductive, inductive, abductive, and competing-hypothesis generation are modes; discriminating predictions and comparison are activated in competing mode."

formulate-hypotheses

Purpose

Generate and refine testable hypotheses from a gap, theory, induction, anomaly, or competing explanations.

When to use / not applicable

Use when a research gap or insight can be stated. Select `deductive`, `inductive`, `abductive`, or `competing-hypotheses`; operationalization may follow any generation mode.

Input contract

mode_contracts:
  deductive: &hypothesis_input
    required: [research_gap_or_observation]
    optional: [theory, anomaly, candidate_explanations, variables, prior_evidence]
    constraints: [at_least_one_observable_consequence]
  inductive: *hypothesis_input
  abductive: *hypothesis_input
  competing-hypotheses: *hypothesis_input

Execution protocol

Do not perform called SOP operations inline; each loaded SOP owns its contract and thresholds.

1. State the gap or observation and select the declared mode. 2. Generate candidate hypotheses without premature filtering using the mode-specific operations below. 3. You MUST load skill `identify-variables` to operationalize the variables. You MUST load skill `specify-relationship` to state their relationships, scope, and boundary conditions. 4. You MUST load skill `score-object` to score each hypothesis against the declared evidence and rubric. If the resulting hypotheses need a focused falsifiability review, consider `falsifiability-audit`. If the strongest claim is ready for a decisive truth-seeking program, `falsification-first-audit` may be the better next tactic.

Mode branches

• `deductive`: derive predictions from an existing theoretical framework. You MUST load skill `identify-theory` to identify the governing theory. You MUST load skill `extract-causal-structure` to derive its causal structure.
• `inductive`: generalize a pattern where theory is weak or absent. You MUST load skill `extract-empirical-regularity` to extract the recurring empirical pattern.
• `abductive`: explain a precisely described anomaly and retain competing explanations. You MUST load skill `characterize-anomaly` to characterize the anomaly. You MUST load skill `generate-competing-hypotheses` to retain distinct explanations.
• `competing-hypotheses`: require mutually discriminating predictions. You MUST load skill `generate-competing-hypotheses` to generate mechanism-distinct alternatives. You MUST load skill `design-discriminating-prediction` to derive discriminating predictions. You MUST load skill `compare-hypotheses` to construct the comparison matrix.

Output contract

mode_contracts:
  deductive: &single_hypothesis_output
    produces: [hypothesis_set, operational_definitions, predictions, falsification_conditions]
    delta_fields: [hypothesis_updates, findings, uncertainties, decisions, open_questions]
  inductive: *single_hypothesis_output
  abductive: *single_hypothesis_output
  competing-hypotheses:
    produces: [hypothesis_set, operational_definitions, predictions, falsification_conditions, comparison_matrix]
    delta_fields: [hypothesis_updates, findings, uncertainties, decisions, open_questions]

Thresholds and quality gates

• Candidate explanation and prediction gates report coverage relative to the declared mechanism/explanation space; record numerator, denominator, batch increment, stopping reason, and source references.
• Base hard gate: at least 1 clear research gap or insight.
• Abductive and competing modes must report relative coverage of candidate explanations and discriminating predictions over the declared mechanism space; fixed counts are not used as a proxy for exhaustiveness.
• Each mechanism must correspond to at least 1 hypothesis candidate.
• A-class source scale is retained as relative tiers over the caller-declared universe: S requires at least 2 structured hypotheses, M at least 3, and L at least 5; deductive theory tiers retain S/M/L at least 2/3/5 named theories; inductive observation tiers retain at least 3/5/8 independent observations; abductive mode requires one precisely defined anomaly. These are minimum source bands, not claims of corpus exhaustiveness.
• Relative audit: declared universe = eligible hypothesis/evidence pool; numerator = qualifying hypotheses, theories, observations, or anomalies; batch increment = newly screened qualifying items per pass; stopping reason = caller rule or saturation decision; source references = source ledger and cited records; direction/threshold rationale = higher coverage is better, with source S/M/L minima preserved as tier gates.

Failure and counterexamples

Reject unfalsifiable wording, variables without operational definitions, and "competing" hypotheses with no observable divergence. Deductive mode is not applicable when no usable theory exists; abductive mode is not applicable without a clear anomaly.

Provenance map

`hypothesis-formulation`, `deductive-hypothesis-generation`, `inductive-hypothesis-generation`, `abductive-hypothesis-generation`, `hypothesis-operationalization`, `theory-mechanism-extraction`, `anomaly-driven-abduction`, `competing-hypothesis-construction`, `competing-hypothesis-matrix`; source thresholds retained.

Legacy context checkpoint / Delta notes

Append candidate list, selected hypotheses, operational definitions, predictions, falsification tests, and unresolved theory conflicts.

Preserved source criteria ledger

| source | source line | kind | source criterion | |---|---:|---|---| | hypothesis-formulation | 41 | textual | ## HARD-GATE | | hypothesis-formulation | 43 | textual | <HARD-GATE> | | hypothesis-formulation | 44 | textual | Preconditions (all must hold before starting): | | hypothesis-formulation | 45 | numeric | 1. A