Consensus

by Manfred Füllsack

WHAT IS IT?

This (simplified) model tests the assumption of producing consensus in prolonged discourses. See below how it works.

Will discussions, if led repeatedly and long enough, eventually result in global consensus? And if so, what conditions are influencing this process?

Motivated by an assumption of Jürgen Habermas and inspired by simulations of Robert Axelrod about disseminating culture and of Joshua Epstein about thoughtlessness (see references), this (simplified) model tests the influence of the operating range on the success of "communicative actions".

HOW IT WORKS

Pushing the <setup>-button generates a set of differently colored agents (small squares). Each color corresponds to a certain conviction which the agent owns in the form of a binary code. This code is part of the agent's permanent endowment and might for example take the form [1 0 0 1]. (For making the model run in browsers and for attributing each individual code a distinct color, length and quality of the code have been reduced to a minimum. The behaviour in question however, has been tested with codes up to a length of 15 and a variation range of digits of 15.)

If you push <go> the agents start "communicating". In each step, agents choose randomly one of the other agents in their surroundings. The radius of this surroundings can be set with the <initial-radius>-slider. A radius of one includes the agents to the North, West, South and East of the acting agent. A radius of two includes the eight adjacent agents plus the agents to the North, West, South and East of this eight neighbors, and so on. If the code of the chosen agent differs from the code of the acting agent, one digit in the code of the chosen agent is randomly selected and replaced with the corresponding digit in the code of the acting agent. By and by, agents, while communicating, will change their convictions and might eventually "agree on a consensus". (Note that in this model, differently to the one of Axelrod, the chosen discourse-partner is the one whose code is changed. Also discourse-partners are chosen randomly, and not with a probability in respect to their similarity)

In the initial setting of the model, each "communicative action" changes just one digit of the code. The Habermasian Theory of Communicative Action however, assumes an "ideal discourse situation" in which, among others, time is not an impediment for discourses. Debaters can discuss untill reaching consensus. By switching on the <boost-all>-switch, you can simulate this "ideal discourse situation". Each communicative action now will result in replacing all digits of the code of the chosen agent. However, they might be replaced again by an other agent in the next moment of time.

Alternatively, with the <boost-majority>-switch you can enhance the discoursive power of just those agents whose code (color) is the most common at the given moment. This switch therewith rather simulates a Matthew-effect. Agents with authority or with otherwise more attention than others will have better chances to disseminate their convictions. (For more information on Mathew-effects see also the Trade and the Lock-in-model)

The <thoughtless>-switch implements an interesting variation that has been inspired by a model of Joshua Epstein. Agents might appease with the success or the current position of their own conviction when surrounded by neighbors of the same conviction. They might become "thoughtless" about different convictions and not so much interested in debates anymore. And they might be motivated to increase their radius of communicative action when sensing dissent or opposition. Thus, by switching on <thoughtless>, agents start testing their initially set neighborhood. When finding agents with different convictions in this neighborhood, they will increase their range of action successively up to the maximum range of the grid. And they will decrease it again when all agents in this range are of the same conviction. The range of action thus can be considered adaptive. The model "learns" how much communicative action is needed. Unfortunately, the computational power in applets restricts the size of the grid here to a maximum that does not demonstrate the differentiation of regions of different action ranges very well. However, the <Radius>- and the <Radus-distribution>-plotters give some picture of how the agents expand and constrain their range of action.

Literature:

Axelrod, Robert (1997): The dissemination of Culture. A model with local convergence and global polarization; in: Journal of Conflict Resolution 41, p. 203-226.

Epstein, Joshua M. (2001): Learning to be Thoughtless: Social Norms and Individual computation; in: Computational Economics 18/1, p. 9-24.

Habermas Jürgen (1981): Die Theorie des kommunikativen Handelns. Frankfurt/M. (The Theory of Communicative Action).