ϵ-Machine and ϵ-Transducer Analysis of Functional Differentiation in Ant Collectives

Area of Science:

• Behavioral Ecology
• Computational Biology
• Animal Collective Behavior

Background:

• Genetically homogeneous animal collectives can display functional behavioral differentiation.
• Symbolic dynamics frameworks like ϵ-machines and ϵ-transducers offer tools to analyze complex behavioral patterns.

Purpose of the Study:

• To investigate functional behavioral differentiation in the ant Pristomyrmex punctatus.
• To model and compare the computational structures underlying distinct movement modes (clustering vs. solitary exploration).

Main Methods:

• Long-term tracking of individual ants (Pristomyrmex punctatus) without marking.
• Reconstruction of individual and population-level ϵ-transducers using symbolic dynamics.
• Comparison of predictive accuracy for three models: ϵ-machine, memoryful ϵ-transducer, and memoryless ϵ-transducer.

Main Results:

• Two distinct movement modes were identified: clustering and solitary exploration.
• Solitary exploration was modeled by a deterministic ϵ-transducer, while clustering required a stochastic one.
• An ant's own behavioral history was the most accurate predictor of its future movement at the examined temporal resolution.
• Agent-based simulations reproduced basic statistics but failed to generate stable clusters, suggesting additional mechanisms are at play.

Conclusions:

• Individual behavioral differences arise from variations in traversing a common state space, not distinct generative programs.
• While social input influences transitions between modes, an individual's output history is sufficient for short-term prediction.
• A two-level organization exists: self-sufficient dynamics within modes and environmentally triggered switches between modes.