Mind · Reasoning
Generalize
Inductive pattern extraction
Definition
Mechanism
Pattern Extraction: Given multiple instances, identify shared structure. Replace specific values with variables. State the invariant that holds across all instances. Test: does pattern predict behavior of new instances? Refine until predictive. The canonical form of Induction in sema's reasoning taxonomy — probable conclusions from specific observations. It employs AnalogyBridge to map specific instances to abstract schemata, validating the invariant across the set.
Preconditions
- Set of specific instances
Postconditions
- General principle extracted
Invariants
- The abstract rule covers every instance it was drawn from.
- Drawn from at least `min_instances` instances.
- Tested against instances not used to form it.
Risks
Failure modes
Overfitting
the extracted pattern includes noise or coincidence from the instances it was drawn from.
Over-generalisation
the rule covers the instances and also cases it should exclude.
Relationships
Builds on
References
Mapping to known solutions
Specific to general logic
Infrastructure
System condition at time T
Used by
References
Learn a pattern from near misses: excluded cases that differ from members by one stated feature
Forcing abstraction via resource starvation
Recognizing implicit patterns in existing practice
Deductive application of principles
Reference
Parameters
min_instancesPositiveInteger, [2, 100]- Instances required before a shared structure counts as a pattern.