Imprecise neural computations as a source of adaptive behaviour in volatile environments

Area of Science:

• Cognitive science
• Computational neuroscience
• Human behavior

Background:

• Humans navigate environments with uncertainty and volatility.
• Existing models of adaptive behavior require complex computations, questioning biological plausibility.
• Understanding efficient adaptation in changing conditions remains a challenge.

Purpose of the Study:

• To investigate whether simple, low-level inferences can explain efficient adaptive behavior in volatile environments.
• To challenge the necessity of complex volatility inference in adaptive behavior models.
• To propose and validate a novel model based on uncertainty and Weber law imprecision.

Main Methods:

• Developed a computational model based on low-level uncertainty inferences and Weber law imprecision.
• Compared model predictions against human behavioral data in volatile environments.
• Evaluated against established models, including optimal adaptive models and reinforcement learning.

Main Results:

• Simple, low-level uncertainty inferences can yield near-optimal adaptive behavior.
• The proposed Weber-imprecision model significantly outperforms complex adaptive models.
• Empirical evidence supports the Weber-imprecision model over high-level inference models.

Conclusions:

• Efficient human adaptation in volatile environments may rely on simple, uncertainty-focused computations.
• The Weber law's imprecision provides a parsimonious explanation for adaptive behavior.
• This finding offers a biologically plausible alternative to complex inference models for adaptation.