Computational validity: using computation to translate behaviours across species.

Area of Science:

• Systems neuroscience
• Comparative psychology
• Computational neuroscience

Background:

• Current cross-species and population research often relies on superficial task similarity.
• This approach may not accurately capture underlying information processing mechanisms.
• A need exists for a framework that bridges behavioral observation with computational underpinnings.

Purpose of the Study:

• To propose a new conceptual framework, computational validity, for cross-species and population research.
• To redefine translation based on computational similarity of information processing, not task similarity.
• To highlight the importance of individual-level behavioral analysis in understanding computation.

Main Methods:

• Defining computational validity based on the similarity of information processing strategies and mechanisms.
• Contrasting computational validity with construct validity and circuit validity.
• Emphasizing an individual-level, subject-centric approach to behavioral characterization.

Main Results:

• Computational similarity, not superficial task similarity, is key for valid translation across species and populations.
• Computational validity directly addresses information processing, offering a deeper understanding than construct validity.
• Individual-level behavioral analysis is crucial as different computations can underlie similar tasks.

Conclusions:

• Computational validity provides a robust framework for understanding behavior and information processing across diverse subjects.
• This framework is particularly relevant for psychiatric disorders and their animal models, addressing information processing dysfunctions.
• Future research should focus on characterizing individual computations to advance systems neuroscience and comparative studies.