# Abductive Hypothesis Generation

> Strategy: Inference to the best explanation in the face of anomalies

- Skill: `yogsoth-ai/abductive-hypothesis-generation` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add yogsoth-ai/abductive-hypothesis-generation`
- Raw SKILL.md: https://api.skillmd.com/api/skills/yogsoth-ai/abductive-hypothesis-generation/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: yogsoth-ai (https://skillmd.com/u/yogsoth-ai)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/yogsoth-ai/abductive-hypothesis-generation

---


# Abductive Hypothesis Generation

Inference to the best explanation in the face of anomalies: when an anomalous phenomenon that existing theory cannot explain is observed, systematically generate candidate explanations and select the most plausible one as the hypothesis.

## When to Use

- A clear anomalous phenomenon is observed (a result inconsistent with existing theoretical predictions)
- Existing theory cannot adequately explain a known phenomenon
- One of several competing explanations must be selected as the most worth testing
- The research starting point is "this result is strange, why?"

Not applicable: no clear anomaly, just wanting to explore a new field → use inductive-hypothesis-generation instead.

## Thinking Framework

**Anomaly → Generate candidate explanations → Rank by plausibility → Best explanation = hypothesis**

The core logic of abductive reasoning:

1. **Anomaly**: precisely describe the anomaly — what phenomenon, inconsistent with what expectation, how large the deviation
2. **Generate candidate explanations**: systematically generate all candidate explanations that can account for the anomaly (no premature filtering)
3. **Rank by plausibility**: rank by plausibility — which explanation is most parsimonious, most consistent with known facts, most testable
4. **Best explanation = hypothesis**: select the most plausible explanation as the working hypothesis, retaining the rest as competing hypotheses

**Core principles of abduction**:
- **Occam's razor**: when explanatory power is comparable, prefer the explanation with fewer assumptions
- **Consistency**: the best explanation should not contradict other known facts
- **Testability**: the best explanation must be able to produce observable predictions (otherwise it cannot be verified)
- **Generation completeness**: candidate explanations must be exhausted before ranking, to avoid premature convergence

## Budget Gate

| Tier | Anomaly description | Candidate explanations | Hypothesis output | Competing hypotheses |
|------|---------|---------|---------|---------|
| S | 1 precisely described anomaly | ≥2 candidate explanations | 1 best-explanation hypothesis | ≥1 competing hypothesis retained |
| M | 1–2 anomalies | ≥3 candidate explanations | ≥2 structured hypotheses | complete plausibility ranking |
| L | ≥2 related anomalies | ≥5 candidate explanations | ≥3 structured hypotheses | complete ranking + discriminating prediction design |

## Default Reference Flow

1. Call the `anomaly-characterization` SOP: precisely describe the anomaly (phenomenon, expectation, deviation, excluded trivial explanations)
2. Call the `explanation-generation` SOP (via the `anomaly-driven-abduction` tactic): systematically generate candidate explanations (no premature filtering)
3. Call the `plausibility-ranking` SOP: rank candidate explanations by parsimony, consistency, and testability
4. Call the `falsifiability-check` SOP: generate a falsification scenario for the best explanation, confirming its testability

## context-checkpoint

Record after each round:
- Anomaly description (precise version, with deviation quantification)
- Candidate explanation list (including excluded trivial explanations and exclusion reasons)
- Plausibility ranking result (including ranking basis)
- Best-explanation hypothesis + competing hypothesis list
- Discriminating predictions (what experiment can distinguish the best explanation from competing explanations)

<!-- BEGIN available-tables (generated) -->

## Available Tactics

Optional, no fixed order; the final leaf is always a sop.

| Tactic | When to use |
| --- | --- |
| anomaly-driven-abduction | Tactic: Inductive/abductive path — describe anomalous phenomena, generate candidate explanations, rank by plausibility |

## Available SOPs

Optional, no fixed order; the final leaf is always a sop.

| SOP | When to use |
| --- | --- |
| falsifiability-check | SOP: check whether a hypothesis meets the falsifiability criterion |

<!-- END available-tables (generated) -->