A Reputation Game Simulation: Emergent Social Phenomena from Information Theory

Here, the data underlying the article "A Reputation Game Simulation: Emergent Social Phenomena from Information Theory" (https://doi.org/10.1002/andp.202100277) is provided.

The data is structured according to the figures it has been used for. There are

• example simulations with basic communication strategies in the folder "single_simulations_3_agents" (Figures 4,5,8,D1)
• statistical simulations with 3 agents and special communication strategies in the folder "statistical_simulations_3_agents" (Figures 9-13, the upper panel of figure 15, figures 16-18, D2 and the left panels of figure D3)
• statistical simulations with 4 agents and special communication strategies in the folder "statistical_simulations_4_agents" (Figure 14, the middle panel of figure 15, the middle panels of figure D3 and the upper panels of figures D4, D5)
• statistical simulations with 5 agents and special communication strategies in the folder "statistical_simulations_5_agents" (The lower panel of figure 15, the right panels of figure D3 and the lower panels of figures D4,D5)
• propaganda simulations in the folder "propaganda_simulations" (Figure 7)

Each simulation is represented by a .json file in which all events that happened during the simulation are collected. Generally, there are three types of events: communications, self-updates (information that the speaker gained about itself is processed) and updates (information that the receiver gained about the speaker and the topic is processed). Additionally, the first line specifies the parameters of each simulation, and the last few lines summarize the final status of the simulation. In the following all important abbreviations are explained:

• parameters
  • decpeting: whether or not agents in generally make dishonest statements
  • listening: whether or not agents in listen to their communication partners
  • disturbing: whether or not agents are particularly risk-taking when making dishonest statements
  • x_est: intrinsic honesties of the agents
  • RSeed: the used random seed
  • NA: number of agents
  • NR: number of rounds
• communication
  • a: speaker
  • b: receiver
  • c: topic
  • J: transmitted message in the form of
• self_update
  • id: number of agent who is updating knowledge about itself
  • Nl, Nt: number of dishonest/honest statements the agent has observed from itself so far
  • I_<id>: knowledge that the agents has about itself after the update in the form of
• update
  • id: number of agent who is updating its knowledge
  • I_<id1>: knowledge that the updating agent has about agent <id1> in the form of
  • Jothers_<id1>_<id2>: last statement that the updating agent heared agent <id1> make about agent <id2>
  • Iothers_<id1>_<id2>: what the updating agent believes that agent <id1> thinks about agent <id2> after the update
  • Cothers_<id1>_<id2>: what the updating agent believes after the update that agent <id1> wants it to think about agent <id2>
  • new_friends/enemies: id of the agent, the updating agent after the update considers a friend/enemy
  • new_K: normalized surprise the updating agent experienced in the last communication (used to calculate kappa)
  • kappa: median of the last ten normalized surprises the updating agent experienced
• final_status
  • id/name: number if the described agent
  • x: the agent's honesty
  • I: the agent's knowledge about all others
  • Nc/Nt/Nl: total number of conversations/honest statements/dishonest statements the agent has made
  • K: the last 10 normalized surprises the agent experienced
  • kappa: the median of K
  • friends/enemies: list of the agent's friends/enemies
  • Jothers/Iothers/Cothers: same as above, now as full array, i.e. the combined information about all others
  • openess/mind/decepting/strategic/egocentric/deceptive/flattering/aggressive/shameless/disturbing: the agent's character traits