Agent-based model for exploring optimal dating strategies

Absztrakt (kivonat)

I formulate and simulate an agent-based model to explore optimal strategies within a dynamic dating market. The model aims to replicate real-world interactions through agents’ attributes, preferences, utilities, and behaviors, as grounded in relevant literature. The agents have incomplete information about their own preferences and the attributes of others. They learn to adjust and refine their own preferences over time based on the feedback (utility) received from interactions. The interactions happen in a 2-dimensional space, the incorporation of the spatial element simulates proximity-based preferences and homophily. With the use of reinforcement learning (RL) algorithms I simulate agent behaviors and evaluate their performance using the complete information stable equilibrium point (Gale-Shapley) as benchmark. This evaluation reveals optimal stopping points where the expected utility of maintaining a current relationship exceeds potential benefits derived from new encounters. I find that attractive people have significantly shorter waiting times before securing partners, a phenomenon amplified by the network effects associated with the ratio of attractive individuals in the population. My model achieves, on average 71.1% of the welfare realized by the Gale-Shapley algorithm. However, the personality traits do not have predictive power regarding dating strategies in my model. These findings contribute to understanding strategic behavior not just in dating markets, but also offer implications for the economic theory of two-sided markets.