The Emergence of Stimulus Relations: Human and Computer Learning.

Area of Science:

• Behavior Analysis
• Artificial Intelligence
• Cognitive Science

Background:

• Traditional stimulus equivalence research relies on human participants.
• Connectionist models (CMs) are emerging as computational analogues for human learning.
• Existing CMs like RELNET simulate stimulus relations using matching-to-sample procedures.

Purpose of the Study:

• To provide an overview of connectionist models in behavior analysis.
• To explore the application of CMs in simulating derived stimulus relations.
• To highlight the symbiotic potential between human and simulated investigations.

Main Methods:

• Review of existing connectionist models (e.g., RELNET, compound stimuli algorithms).
• Discussion of derived stimulus relations in human academic remediation.
• Introduction of a working example: Emergent Virtual Analytics (EVA) neural network.

Main Results:

• CMs demonstrate the capacity to approximate human acquisition of stimulus relations.
• Neural networks learn to derive novel or untrained stimulus relations over training epochs.
• EVA model illustrates how CMs can simulate and predict human-derived stimulus relations.

Conclusions:

• Connectionist models offer a valuable tool for understanding, simulating, and predicting human learning.
• The symbiotic relationship between human and artificial investigations enhances behavior analysis.
• Emerging neural network approaches like EVA advance practical and experimental behavior analysis.