Belief propagation in multi-agent communities (IV)

Roles

In this section, we introduce two different agent roles: credulous and sceptical agents. Credulous agents change their words more easily than the normal ones, the contrary happens with the sceptical. The probability for an agent to change its word in an encounter with another is:

p(change) = c*otherBelief/(ownBelief+otherBelief)

where ownBelief is the belief of the agent in its own word, and otherBelief is the belief of the other agent in its word. c is a constant for each agent that determines the degree of credulity of the agent and can have the following values:
For credulous behaviour, the probability of not changing its word is:

p(keep) = k*ownBelief/(ownBelief+otherBelief)

where k is a constant for each agente, with a value between 0 and 1. If k is zero, the the agen always changes its word, and if k is less than 1, the agent is credulous. In the first set of simulations, we have experimented with scepticism. In this experiments the word uniformization takes a long time to appear, and in some cases it can't be achieved. For low values of c with 5% of the agents being sceptical (i.e. 5% of them have a value of c between 0 and 0.5, the rest have a value of 1 for c), uniformization may not be achieved in any of the words, as is shown in the next simulation.

This is the complete OOCSMP code.

The belief indicators curve exhibit an interesting behaviour in these experiments. It reaches lower values than in previous cases, and grows very little at the beginning, but after some time, begins to grow faster. This moment is associated with a decrease of the mean deviation of the words.
Credulous agents exhibit a similar behaviour. If few of the total population are little credoulous, the uniformization process is slowed, but uniformization is reached. On the other hand, if some of them are strongly credulous, uniformization may not be reached.
As an example of this phenomenon, the next applet shows a simulation of 400 agents, 50% of which were strong credulous (50% of them had k between 0 and 0.5, for the rest the value of k was 1). It can be observed that the belief indicators curves exhibit a similar behaviour than with sceptical agents, although the value reached is lower than in the previous cases. It grows very little at the beginnig of the simulation, and later, it grows faster. The moment of the change is associated with a strong displacement of the deviations.

We have also made experiments with all the population at different positions in the range from credulous to sceptical. No uniformization is reached in these cases.

Conclusions, future work

We have presented a model that simulates the propagation of a vocabulary naming movements in a multi-agent environment. The agents have belief indicators associated to each word. Agents trust their own word to a degree depending on these indicators. Several trusting and moving strategies are presented. Vocabulary uniformization is in most simulations, but moving strategies accelerates the process. Agent roles are also examinated, credulous and sceptical agents make the uniformization process to be slower, and sometimes a stationary state with no uniformization emerges.
These simulations can model the way beliefs are transmitted in a closed group of people. In a situation where people have different beliefs none of which is a priory better, a unique belief spreads and sometimes covers all the population individuals. Strong sceptical or credulous agents prevent a single belief from spreading in the whole population, and a stable situation with several coexistent belief arises.

In the future, we plan to investigate some other movement strategies, such as agent moving to the position where more individuals share the same word. This movement would result in a reinforcing of the agent beliefs. Other roles, such as lying or non-communicating agents can also be investigated.
We are also considering to add some artificial intelligence logic to our simulation language, in order to give agents a more sophisticated way of reasoning.

Belief propagation in multi-agent communities (III)

References