Inductive reasoning about causally transmitted properties.

Area of Science:

• Cognitive Science
• Computational Modeling
• Psychology

Background:

• Intuitive theories guide human inferences in various contexts.
• Previous models often lack context-specific reasoning capabilities.
• Understanding category-based induction is crucial for cognitive modeling.

Purpose of the Study:

• To present a novel computational model formalizing intuitive theories as probabilistic processes.
• To investigate context-sensitive reasoning in category-based induction, particularly for causally transmitted properties.
• To compare the predictive power of causal versus taxonomic models in human reasoning.

Main Methods:

• Formalized intuitive theories as probabilistic processes over structured representations.
• Developed a computational model for category-based induction of causally transmitted properties.
• Conducted three experiments with human participants using taxonomic and food web knowledge, including artificial and novel food webs.

Main Results:

• Human inferences about causally transmitted properties align with the computational model's predictions.
• The model demonstrated strong fits for reasoning about artificial food webs.
• A double-dissociation was observed: the causal model predicted disease inferences, while a taxonomic model predicted gene inferences, highlighting distinct reasoning mechanisms.

Conclusions:

• The proposed computational framework effectively models context-sensitive, causal reasoning in category-based induction.
• Human reasoning employs distinct causal and taxonomic strategies depending on the domain (e.g., diseases vs. genes).
• Further research is needed to develop comprehensive models of context-sensitive intuitive reasoning.