Metacognition as a Consequence of Competing Evolutionary Time Scales

Area of Science:

• Evolutionary biology
• Computational neuroscience
• Machine learning

Background:

• Coevolving systems exhibit complex interactions and adaptation across multiple temporal scales.
• Understanding the mechanisms of multi-level adaptation in evolution remains a challenge.

Purpose of the Study:

• To formalize evolutionary multi-scale processing as a form of metacognition.
• To investigate the conditions under which metacognitive systems evolve.
• To demonstrate the energetic efficiency and prevalence of metacognition in evolutionary processes.

Main Methods:

• Defining metaprocessing and metacognition in the context of evolutionary systems.
• Analyzing fitness landscapes that vary on multiple time scales.
• Examining various modeling approaches like active inference networks, predator-prey models, genetic algorithms, and generative adversarial networks.
• Proving the energetic efficiency of metacognition compared to object-level cognition.

Main Results:

• Metacognitive systems are expected to evolve when fitness landscapes vary across multiple time scales.
• Multiple time scales emerge in coevolutionary processes with complex interactions.
• Metacognition is energetically more efficient than object-level cognition under multi-timescale selection.
• Existing coevolutionary models demonstrate emergent multi-timescales indicative of metacognition.

Conclusions:

• Multi-scale processing is a necessary component for adaptation in many evolutionary scenarios.
• Metacognitive systems are a prevalent and vital aspect of (co-)evolution.
• Coarse-grained structures naturally emerge in resource-limited systems, supporting metacognitive evolution.