Mind · Strategy
HypothesisLadder
Bayesian belief updating via falsification rungs
Definition
Mechanism
The agent explicitly lists its current hypotheses about the world state and assigns probabilities. As new data arrives, it updates these probabilities using BayesUpdate. It acts on the highest-probability Hypothesis but keeps others alive. It structures Abduction into falsifiable rungs, climbing to higher certainty only when an Experiment validates the current level.
Preconditions
- Data is valid
Postconditions
- Experiment defined
Invariants
- Ascension Rule: Cannot move to Hypothesis(N+1) until Hypothesis(N) is validated by evidence.
- Exclusivity: Hypotheses at the same rung must be mutually exclusive (required for Bayesian probability assignment).
- Falsifiability: Each rung must have a testable disprove condition.
Risks
Failure modes
Clinging to low-probability priors.
Relationships
Builds on
Composes with
Structured causal discovery via variable isolation
References
Inference to the best explanation — provisional, ranked by simplicity/scope/coherence
Mathematically rigorous belief revision
Infrastructure
Testable prediction staged for falsification attempts — provisional, not asserted
Nothing in this library builds on this pattern yet.