An Agent-Based Model of Collective Decision-making in Correlated Environments

In many complex systems, from robot and insect swarms to human social systems, agents take decisions collectively, using information retrieved from the environment and from each other. This information is usually correlated to some
extent – e.g., voters reading the same media outlets, animals receiving the same cues from their environment, or people listening to the same opinion leaders. Taking inspiration from human social systems, we consider the case of collective
decision-making between two choices, one being the correct one. We break down the problem of collective decision making in correlated environments into two components: (i) how likely different configurations of information environments
are to show the correct option and (ii) how likely different configurations of collectives are to detect the majority shown by the environment. An agent-based model is presented, where agents scan an information environment, composed
of correlated and uncorrelated sources, and form individual opinions based on the information perceived. Once individual opinions are formed through a majority function, agents take a majority vote to determine their chosen option.
Preliminary results show how different population parameters lead to different decision accuracy in similar information environments, and how the two steps of opinion formation and collective vote can skew the collective’s perception of the environment positively or negatively. Future work will expand these results by allowing agents to form local groups before taking decisions collectively.