"Sociophysicists" provoke sociologists: new computer simulation models explain the mechanism which makes moderate individuals radicalise. It is sufficient to be a little self-confident and to keep contact with a few extremists to let extremists win a majority in society.

Keywords:

Extremist Attitudes; Opinion Dynamics; Sociophysics

1.1

"German conditions" is the title of a frightening study conducted by the Bielefeld Institute of Interdisciplinary Research into Conflict and Violence: people hostile against foreigners, Jews, homosexuals, handicapped and homeless people have the potential to become a majority. It is not a nice observation that Die Zeit published last Thursday. Readers ask themselves uneasily whether and to what extent the centre — however one could describe the centre — is moving to the right.

1.2

A week before, the Journal of Artificial Societies and Social Simulation (JASSS) published a study (Deffuant et al 2003) on the conditions in which extremist attitudes win a majority. Here, new hope arises: it might be fine if social scientists could give us an answer!

1.3

The method of the French authors is computer simulation. This is not unheard of in the social sciences. As early as the 1960s, modellers such as Dennis Meadows designed complex systems of equations to represent global interactions. Their mathematical models grew so complicated that conclusions could only be drawn by testing their behaviour on large mainframes. This is how the Club of Rome report on The Limits to Growth was compiled in 1972.

1.4

From then on, beside a barely verbal social science, and critically viewed by it, a new branch of research was established which tries to understand social processes by way of computer simulation. The JASSS paper is nevertheless uncommon, beginning with its authors: the paper was written by a team of researchers led by Guillaume Deffuant, of Cemagref, in the framework of a European project devoted to the attitudes of agriculturists towards the environment. The project also involved the physicist Gérard Weisbuch, of the Ecole normale supérieure (ENS).

1.5

What, please, do physicists do in the sphere of the social sciences?

1.6

Understanding. And they introduce methods which might nevertheless be appropriate for the social world as a target system. This is because physical phenomena can be explained as results of the interactions of countless elements — and what could society be, if not the interactions of many elements, although these elements are not atoms but human beings?

1.7

"Sociophysics" is the name of a new research tradition building on this insight. Typically, in their computer simulations they have very simple elements interact with each other. These represent the members of a society and are much simpler than the "intelligent agents" of "artificial societies" whose colourful pixel diagrams have been introduced to magazines and the TV for several years.

1.8

The model presented in the JASSS paper describes individuals who have only two properties: first, an opinion about something: foreigners, music, religion, science, cheese, no matter what. This opinion can be represented as a point on an axis from -1 to +1: these are the extremes, while 0 is the centre. The second property is uncertainty, also expressed by a number. The law governing the interactions is the following. The individuals' opinions and uncertainties influence each other the more the more similar they are and the individual who is more certain exerts the greater impact on the one who is less certain. If the two individuals have the same degree of certainty, they will not influence each other at all.

1.9

That's all, now it's the computer's turn! In every simulation step, the program picks two individuals at random, compares their opinions and certainties and updates their values accordingly. While this happens several hundred or several thousand times, the researchers observe the computer screen which shows them how the opinions of the simulated individuals change. The certainties — which, too, are changed — are represented in different colours (see figures). The authors played with all parameters. The most important parameters were the proportion and distribution of the extreme opinions at the outset of the simulation and the initial values of uncertainty and their distribution.

1.10

The most interesting run was of course the one in which first a majority is found in the centre which then moves upward, toward one of the extremes (figure 1). Here, in the beginning the extremists on either side were equally strong, but the uncertainty of the non-extremists was very high, which is why the situation was unstable such that a small asymmetry among the moderates or the random picking of pairs of individuals had a devastating effect.

Figure 1. Example of single extreme convergence. Horizontal axis: iterations. Vertical axis: opinions. Coloured axis: uncertainties. pe = 0.1, u = 1.4, µ = 0.5, delta = 0, ue = 0.1, N = 200. The majority (98.33%) of initially moderate agents (initially green, between the two extremes) is attracted by the negative extreme. The convergence indicator value is 0.97. From Deffuant et al (2003)

1.11

A uncertain centre is more susceptible to extremists. This sounds plausible, but not sensational. But as compared to the so-called "extremism research" which more or less has the character of secret services, the French paper must be evaluated at least as having some content.

1.12

In several simulations the realistic outcome was that extremists keep to themselves if they do not have enough neighbours and if the moderates are the more certain individuals. Radical minorities that became majorities have indeed often occurred in history. But they had a chance only if their ideas — such as in the case of Christianity — were able to convince or persuade others, i.e. if there were others with similar opinions.

1.13

What is less plausible is the paradigm according to which neighbourhood always leads to agglomeration. Particularly at the extreme wings, slight differences in opinions can have dramatic affects: the history of communist parties or the hip-hop movement give marvellous examples.

1.14

For Rainer Hegselmann, philosopher at the University of Bayreuth, Germany, who has carried out computer simulations for many years, the model presented in JASSS has additional weaknesses. "Why should the centre be less certain than the extremists? Is it not just the centre which is so very self-convinced? Extremists could also be modelled in a more realistic manner. It is not the certainty which is typical for extremist positions, but the one-sidedness of the opinion itself: people will be more easily influenced by those who are even more extreme than by moderates."

1.15

The criticism of Petra Ahrweiler, senior researcher at the University of Hamburg and one of the major hopes of German social science, goes even further. For her, the social models of the physicists are, above all, too simple. "The relative difference of opinions is expressed in just one number in this model. Reality is more intricate." But aren't models always simpler than reality? "Well, yes", answers the sociologist experienced in simulation, "but nowadays richer models can be designed. We can endow individuals with more knowledge and with more complicated rules of behaviour. Under which conditions will an opinion be influenced? Which contents are important? What is the relation between opinions, needs, and interests? How convincing is an individual? — and so on." This is the "intelligent agent approach" quoted in the first paragraphs, which is going to be the leading idea of the community, as Klaus Troitzsch, Koblenz-based pioneer of social simulation, puts it.

1.16

The debate about the best method in social simulation is far from being closed. Ahrweiler and Hegselmann as well as the authors of the JASSS report and other representatives of the social simulation community travelled to the Bielefeld Centre for Interdisciplinary Research (ZiF) last summer to discuss "Sociophysics". There, two cultures clashed, reports Petra Ahrweiler. The physicists and their philosophical companions prefer models with primitive particles. Why be complicated if there are simple solutions? Sociologists on the contrary rather like intelligent agents. Why not make use of today's computer power to conceive models nearer to reality?

1.17

But there are also adversaries of the extremism model among the representatives of their own camp. Wolfgang Weidlich, professor emeritus of physics in Stuttgart and a passionate social modeller, could not match his experience of the Nazi times with the model. But the discussion was not continued far enough.

1.18

"Intelligent agents", endowed with much individuality, are currently one of the two methodological extremes in the modellers' community. Brownian social particles, stupid like sand, form the other extreme. Which position will prevail? Or will the two extremes meet in the centre? Perhaps our French experts in the simulation of extremism should take an interest in this opinion formation process, too.

1.19

Anyway, for our pressing problem — the German conditions — the authors' work yields a lot: we must not only worry about the bare existence of anti-minority attitudes, but also about the fact that there are people with opinions similar to the extremists — and about the fact that so many Germans are uncertain about their opinions.

---

See the response to this article by G. Deffuant, G. Weisbuch, F. Amblard and T. Faure (2003) 'Simple is beautiful ... and necessary' Journal of Artificial Societies and Social Simulation vol. 6, no. 1 <https://www.jasss.org/6/1/6.html>

Reference

DEFFUANT, Guillaume, Frédéric Amblard, Gérard Weisbuch and Thierry Faure (2002) How can extremism prevail? A study based on the relative agreement interaction model Journal of Artificial Societies and Social Simulation vol. 5, no. 4 <https://www.jasss.org/5/4/1.html>

Return to Contents of this issue

© Copyright Journal of Artificial Societies and Social Simulation, [2003]