Autor: A. T. S

Afiliação: Universidade Eduardo Mondlane, Departamento de Matemática e Informática, Moçambique, Maputo, arnaldo.t.siote@uem.ac.mz or tsakane07@gmail.com

Contextualization

Understanding human learning and cognition has evolved through traditional psychological theories: behaviorism, cognitivism, and constructivism. Behaviorism emphasizes observable behaviors. Cognitivism highlights internal mental processes. Constructivism posits that knowledge is actively constructed through social interactions. Despite their contributions, these theories struggle to capture the full complexity and dynamism of cognition.

Recent advancements in artificial intelligence provide novel computational models that can extend and test these frameworks. This study reveals how computational models both validate and illuminate psychological understanding of cognition.

Objectives

Investigate the intersection of artificial intelligence, neuroscience, and psychology by demonstrating artificial intelligence’s capacity to simulate cognitive processes; providing new insights into learning mechanisms; developing theoretical extensions that bridge disciplines, addressing interdisciplinary challenges; evaluating broader implications for cognitive science.

Methodology

We conducted a systematic literature review following PRISMA guidelines, screening 51 publications across psychology, neuroscience, and artificial intelligence. Our analysis comprised four phases: literature review, development of a comparative, multidimensional framework, case-study of three representative artificial intelligence implementations, and theoretical integration to link computational models and psychology.

Findings

Our findings reveal how artificial intelligence operationalizes psychological learning theories: DeepMind's reinforcement learning algorithms provide precise implementations of behaviorist conditioning principles with reward‐driven updates; GPT-4's transformer architecture operationalizes cognitivist concepts through attention mechanisms; multi-agent systems formalize constructivist principles through interactive, collaborative agents.

Conclusion

This interdisciplinary study demonstrates bidirectional benefits: computational implementations provide mechanistic explanations for psychological phenomena, while psychological theories guide more cognitively plausible artificial intelligence development. We identify critical research gaps in modeling consciousness, embodied cognition, and emotional intelligence, and we outline a path toward a unified framework for understanding and engineering cognitive processes.

Keywords: learning theories, neural networks, computational neuroscience.