Skip to content

Population & Preferences

In this model, preference groups define ordinal color rankings, while individual agents additionally carry their own preference distribution over those rankings.

Implementation Reference

Population initialization and model-level preference setup are documented in ParticipationModel. Agent-level preference construction and knowledge updates are documented in VoteAgent.

Main Knobs

  • num_agents (>= 1): number of voting agents
  • known_cells (>= 0): number of cells sampled per step for local estimation
  • num_personality_groups (>= 1): number of preference-group orderings (bounded by factorial(num_colors))
  • personal_preference_peakedness (> 0): concentration of personal_opt_dist
  • initial_agent_assets (>= 0): identical starting assets for all agents

Semantics

Preference groups are permutations over color IDs. They determine the order in which colors are preferred, but not how strongly they are preferred.

personal_opt_dist is then generated to follow each agent’s preference-group ordering while still allowing variation in preference intensity across agents.

Knowledge sampling via known_cells determines how much local information an agent has when constructing her current estimate for voting.