Mind · Reasoning
Abduction
Inference to the best explanation — provisional, ranked by simplicity/scope/coherence
Definition
Mechanism
Inference to the best explanation. Given a surprising observation, generate candidate explanations and Rank them by simplicity, scope, and coherence with existing knowledge. Adopt the highest-ranked as working Hypothesis, flagged as provisional — not proven. Uses ChainOfThought to trace the reasoning path from observation to explanation so the inferential leap is explicit and verifiable. The classical third mode of reasoning alongside Deduction (necessary) and Induction (probable from instances): Abduction is probable from a single anomaly plus background knowledge.
Preconditions
- Set of observations, at least one unexplained by existing model.
- No deductive path to the explanation.
Postconditions
- Hypothesis generated and ranked by likelihood.
- Working hypothesis adopted with provisional status.
Invariants
- At least two candidate explanations are compared.
- The explanation covers every observed anomaly in the input.
- Occam's Razor decides between candidates of similar scope.
- The output is flagged provisional, never asserted as proven.
Risks
Failure modes
Conspiracy Thinking
Preferring complex, coherent explanations over simple, messy ones (overfitting the narrative).
Stopping too early
adopting the first plausible explanation without considering alternatives.
Treating the provisional hypothesis as proven in downstream reasoning.
Relationships
Builds on
Composes with
Step-by-step reasoning (Macro for Think(Chain))
References
General to specific logic
Infrastructure
Testable prediction staged for falsification attempts — provisional, not asserted
Specific to general logic
Infrastructure
Order items by score
Used by
References
Bayesian belief updating via falsification rungs
Reference