Belief propagation in multi-agent communities (III)

Trusting strategies

In this set of experiments, we have added non-deterministic behaviour (or 'noise') to the way agents trust each other. When several agents shares the same position, each agent 'talks' to the rest in turns as before, but an agent is 'convinced' in a non-deterministic way. When an agent is listening to the 'talking' agent, it has a probability of being convinced, computed as otherBelief/(ownBelief+otherBelief). In this experiments, the uniformity takes longer to appear.

The next simulation shows a simulation with 400 agents, using this trusting strategy.

It can be observed that with non-determinism, words that take longer to reach uniformity, are associated with lower beliefs, in opposition to the behaviour of communities of deterministic agents.

This model is autocatalitic, belief indicators always grow. To avoid this, and to make the model a bit more realistic, we have modified it to make belief indicators decrease when an agent shares the same position with another with a different word associated to the same movement, that is not convinced by its word. Some acceleration in the uniformity is obtained. In this way, lonely agents with words different from its neighbours are quickly 'convinced' by them. Even in this case, the curves for belief indicators exhibit growing behaviour.

Movement strategies

In previous experiments, agents didnīt have any moving strategy, their movements were completely random. In this section, we consider two different kinds of movement strategies, that try to improve the speed of uniformization.

In the first strategy the agent choice is proportional to its belief in the word associated to each movement, i.e. the higher the belief in a word with respect to the others, the higher the probability to choose that movement. This strategy accelerates notabily the uniformization time. For example, the next is a simulation of 400 agents using this strategy.

This is the complete OOCSMP code.

The uniformization time of this simulation can be compared with that of simulation 1 (deterministic case, no strategy). It can be noted that the belief curves are stratified. The higher belief value corresponds to the first word that reaches uniformization. These is due to the fact this movement has been selected more frecuently.

In the second strategy the agent chooses the movement associated to the word with less confidence (in a draw case, a random word is chosen). After moving, the agent requests the associated word to the agents sharing the same position.

The next model shows a simulation of a community of 400 agents that uses this strategy. In this simulation, it can be seen that the beliefs associated with all the words grow at the same rate. This is due to the fact that each agent always asks the word associated with its lower belief, thus beliefs for each agent remain equal.

This is the complete OOCSMP code.

This strategy seems to be better than the basic one, but a little slower than the previous one, although it mantains belief indicators for each word at similar values.

Belief propagation in multi-agent communities (II)
Belief propagation in multi-agent communities (IV)