Sensory perception relies on fitness-maximizing codes.

Area of Science:

• Neuroscience
• Computer Science
• Economics
• Computational Neuroscience

Background:

• Sensory information is typically assumed to be as accurate as biological limits permit.
• However, perfect accuracy may not always enhance survival chances.
• This challenges the traditional view of sensory encoding.

Purpose of the Study:

• To investigate if sensory encoding strategies are optimized for fitness maximization rather than pure accuracy.
• To develop a theoretical framework integrating neuroscience, computer science, and economics for fitness-maximizing encoding.
• To explore how environmental pressures shape sensory processing.

Main Methods:

• Developed a unified normative framework for fitness-maximizing encoding.
• Conducted behavioral experiments with human participants.
• Utilized deep neural networks with information capacity constraints.
• Analyzed human functional magnetic resonance imaging (fMRI) data.

Main Results:

• Human sensory encoding strategies flexibly adapt to maximize fitness.
• Deep neural networks trained on the same task confirmed these findings.
• Novel behavioral goals induced efficient stimulus representations in early sensory structures.
• Evidence suggests fitness-maximizing rules are applied early in sensory processing.

Conclusions:

• Sensory encoding is not solely driven by accuracy but by fitness optimization.
• Environmental pressures influence sensory processing at early stages.
• This principle applies to both biological organisms and artificial intelligence.