lex-analogical-reasoning

Analogical reasoning engine (Gentner structure mapping, Hofstadter, Holyoak multi-constraint) for brain-modeled agentic AI.

What It Does

Models the cognitive process of analogical reasoning: finding structural correspondences between domains and using those mappings to transfer knowledge, generate predictions, and solve problems in unfamiliar territory. Based on Gentner's Structure-Mapping Theory, which prioritizes relational structure over surface features. Analogies are tracked with strength scores that evolve through reinforcement and decay.

Core Concept: Structure Mapping

An analogy is a set of correspondences between a source domain (familiar) and a target domain (new):

# Create a structural mapping between two domains
client.create_analogy(
  source_domain: :water_flow,
  target_domain: :electrical_circuit,
  mappings: [
    { source: :pressure, target: :voltage },
    { source: :flow_rate, target: :current },
    { source: :pipe_resistance, target: :resistance }
  ],
  mapping_type: :relational,
  strength: 0.8
)

Usage

client = Legion::Extensions::AnalogicalReasoning::Client.new

# Apply an analogy to map a source concept to the target domain
client.apply_analogy(analogy_id: id, source_element: :pressure)
# => { success: true, mapped: true, target_element: :voltage }

# Transfer knowledge across domains
client.cross_domain_transfer(
  analogy_id: id,
  source_knowledge: { high_pressure: :causes_fast_flow }
)

# Reinforce based on application success
client.reinforce_analogy(analogy_id: id, success: true)

# Find productive analogies
client.productive_analogies
# => { analogies: [...], count: 3 }

# Maintenance
client.update_analogical_reasoning

Integration

Wire into reasoning phases where the agent needs to reason about unfamiliar domains by mapping from known domains. Transfer results can be stored as new semantic traces in lex-memory.

Development

bundle install
bundle exec rspec
bundle exec rubocop

License

MIT