Naturalistic Computational Cognitive Science: Towards generalizable models and theories that capture the full range of natural behavior (2502.20349v2)

Abstract: How can cognitive science build generalizable theories that span the full scope of natural situations and behaviors? We argue that progress in AI offers timely opportunities for cognitive science to embrace experiments with increasingly naturalistic stimuli, tasks, and behaviors; and computational models that can accommodate these changes. We first review a growing body of research spanning neuroscience, cognitive science, and AI that suggests that incorporating a broader range of naturalistic experimental paradigms, and models that accommodate them, may be necessary to resolve some aspects of natural intelligence and ensure that our theories generalize. First, we review cases from cognitive science and neuroscience where naturalistic paradigms elicit distinct behaviors or engage different processes. We then discuss recent progress in AI that shows that learning from naturalistic data yields qualitatively different patterns of behavior and generalization, and discuss how these findings impact the conclusions we draw from cognitive modeling, and can help yield new hypotheses for the roots of cognitive and neural phenomena. We then suggest that integrating recent progress in AI and cognitive science will enable us to engage with more naturalistic phenomena without giving up experimental control or the pursuit of theoretically grounded understanding. We offer practical guidance on how methodological practices can contribute to cumulative progress in naturalistic computational cognitive science, and illustrate a path towards building computational models that solve the real problems of natural cognition, together with a reductive understanding of the processes and principles by which they do so.

Naturalistic Computational Cognitive Science: Bridging AI and Cognitive Science

The paper "Naturalistic Computational Cognitive Science: Towards generalizable models and theories that capture the full range of natural behavior" by Wilka Carvalho and Andrew Lampinen addresses the ongoing integration of AI advancements into the field of cognitive science. This integration provides the opportunity to develop more generalizable models and theories that are capable of capturing the complete spectrum of human behavior in realistic contexts.

The authors emphasize the necessity for cognitive science to not only distill cognition into simpler models but also embrace the complexity of naturalistic stimuli, tasks, and behaviors made feasible through recent advances in AI. They review a broad spectrum of research across neuroscience, cognitive science, and AI that proposes the inclusion of more naturalistic experimental frameworks is crucial for resolving aspects of natural intelligence and ensuring theories adequately generalize.

The paper lays out a path for cognitive science to harness AI's advancements by incorporating more naturalistic phenomena without forgoing experimental control or a theoretically sound understanding. The authors suggest that recent progress in AI, such as self-supervised and human-like learning models, provides tools that cognitive science can utilize to adapt experiments and models to more realistic conditions.

Key recommendations from the paper include:

1. Embracing Naturalistic Paradigms: Cognitive experiments should cover a broader range of natural stimuli and conditions, thus increasing ecological validity. This fostered approach promotes a complete understanding of cognition applicable to real-world scenarios.
2. Integrating AI Advances: Cognitive science should leverage modern AI models and methodologies to analyze naturalistic stimuli extensively. This integration allows AI models that perform tasks in naturalistic settings to be insightful test beds for cognitive theories.
3. Developing Generalizable Models: There's a call to develop models that are not just task-specific but flexible enough to perform across multiple tasks seamlessly, similar to many AI models that demonstrate multi-domain generalization.

The authors do not shy away from the complexities involved in this integration. They acknowledge that naturalistic data introduce unique challenges like confounding variables and increased difficulty capturing the mechanisms of cognition. Nonetheless, they see value in reconciling these challenges through innovative methodological practices that capitalize on computational advancements, thereby forging a more comprehensive understanding of natural intelligence.

The implications of this paper are substantial with regard to future directions in both cognitive science and AI development:

• Theoretical Implications: By advocating for a blend of AI rigor with cognitive theoretical foundations, the approach facilitates creating models that predict human cognition and behavior in more naturalistic environments. This paradigm could lead to the refinement and potential overhaul of existing cognitive theories that have traditionally relied on isolated parameters and simplified stimuli.
• Practical Implications: Practically, the better alignment between AI capabilities and cognitive science can lead to the development of technologies and tools that better understand, predict, and enhance human cognition, learning, and decision-making in naturalistic settings.

This paper proposes a forward-looking vision for both fields, suggesting a collaborative platform where AI and cognitive science not only coexist but actively enrich one another, facilitating advancements towards a more comprehensive understanding of intelligence as it manifests in natural settings.