This model simulates how agents (representing individuals) update beliefs on a diverse set of political issues within a social context. Each agent belongs to a group that defines a subjective belief system or cognitive model, which governs how beliefs are interconnected for the group's members. We infer such cognitive models from survey data. Beliefs are represented as vectors in the model and they too are initially inferred from survey data. The beliefs are updated either through self-reflection (agents periodically reflect on their beliefs, introducing some randomness, i.e. noise) or social influence (agents interact with their neighbors, and may adopt aspects of their neighbors' beliefs). However, agents seek to maintain a coherent belief state by rejecting or accepting such new beliefs if this would increase the agent's belief coherence. In particular, the decision to adopt a new belief is probabilistic, based on how the new belief aligns with the agent's existing cognitive model. This constraint can be tuned by a parameter, which represents the importance of coherence for the agents' decision-making.