Induction of relational schemas: common processes in reasoning and complex learning

Area of Science:

• Cognitive Psychology
• Learning Sciences
• Computational Neuroscience

Background:

• Relational schema theory proposes that humans can induce abstract representations of task structures.
• Understanding how these structures are formed from specific instances is crucial for explaining complex learning phenomena.
• Existing theories like configural learning have limitations in explaining certain learning behaviors.

Purpose of the Study:

• To investigate whether participants can induce relational schemas from specific task instances.
• To test predictions of relational schema theory against alternative nonstructural theories.
• To explain learning set acquisition and the reversal shift phenomenon using relational schema induction.

Main Methods:

• Five experiments were conducted where participants learned initial-state, operator, and end-state instances.
• Participants were tasked with predicting the correct end-state given an initial-state and operator.
• Performance was evaluated based on the ability to predict items in new isomorphic problems and transfer learning.

Main Results:

• Adult participants efficiently induced relational schemas, demonstrating the ability to predict unseen items in isomorphic problems.
• Induced relational schemas exhibited properties like omnidirectional access and efficient transfer between isomorphs.
• Structural coherence significantly impacted learning, and the "learning to learn" effect was observed and explained.

Conclusions:

• The findings strongly support relational schema theory, providing evidence for the induction of coherent task structures.
• Relational schema induction and structure mapping offer a comprehensive explanation for learning set acquisition.
• The theory also successfully explains the reversal shift phenomenon, outperforming configural and nonstructural learning models.