Threshold regions governing payoff-sensitivity effects

Investigate whether upper and lower threshold regions exist within the topic-competition model that determine when adjusting the payoff-sensitivity parameter α can effectively improve system performance, and characterize their influence on the dynamics of strategy evolution.

Background

The paper examines an agent-based evolutionary model in which social-media topics extracted by BERTopic compete as strategies. Agents revise their topic preferences through imitation under bounded rationality, while the parameter α controls the relative contribution of an agent’s own topic payoff and the payoff associated with the opponent’s strategy. Experiments using different interaction-weighting configurations produced nonlinear and configuration-dependent effects of α on simulation error. In particular, some α values substantially improved fitting performance under one weighting scheme but not another, motivating the unresolved possibility that threshold regions govern when payoff-sensitivity adjustments are beneficial. The paper explicitly leaves systematic exploration of this relationship and the existence of such thresholds to future research.