Computer Simulations in Science and Technology Studies

Despite the progress made in the last few years using computer models for sociological purposes (e.g. Gilbert and Conte 1995; Klüver 1995; Gilbert and Doran 1994), most sociologists probably still associate the construction of formal models and their implementation as computer programs with the (natural) sciences and as methods which are therefore alien to the mainstream of sociology. This is especially the case with the sociology of science and the related fields of Science and Technology Studies (STS). Although Thagard (1989), for example, has analysed scientific theories with the use of recurrent interactive networks, there are few other examples.

This chapter outlines the history of a growing research community: the “invisible college” (Mullins 1973) of scientists who work on computer simulations in Science and Technology Studies (STS). Their common interest enables at least two possible research areas which are only just emerging.

The formation of causal theories is of central importance both for science and common sense reasoning. Causal knowledge is essential for the explanation, prediction, and control of natural processes. Yet it is often questioned whether there is anything like a rational method for discovering such causes. The use of computer simulations has opened up new possibilities for the systematic study of scientific methodology. Testing a proposed method in a simulated environment allows for better control of relevant aspects, and the scope and limitations of a method stand out more clearly. From a perspective that integrates ideas from philosophy of science and artificial intelligence I argue for a conception which views scientific discovery as a kind of problem solving activity that can be subjected to methodological analysis.

Computer linguistic techniques may be useful as a new, innovative tool in the generation of new conceptual links in biomedical knowledge. We will present results from modelling several cases of discovery in pharmaceutical research. Our approach involves the active reconstruction of disconnected, implicit or hidden logical inference patterns in scientific literature. This approach is attractive, not due to the value of the logical model per se, but because it is or can be linked with specific cultural practices within the scientific community. The following topics related to our modelling of discovery processes in pharmaceutical research will be discussed:

Chernenko (1989) used three coupled differential equations to describe the evolution of technologies in terms of the time dependent sizes of three subpopulations applying hunter/gatherer, agricultural, and industrial production technologies, respectively.

Applying “social simulation” to the field of “social studies of science” is commonly thought to be a contradiction in terms². Simulations have to rely on adequate “representations” of their targets; but this very reliability has been dismissed by social constructivism. This dismissal includes advice on how to avoid shortsighted perspectives: “Any study of mathematics, calculations, theories and forms in general should […] look at how observers move in space and time, how the mobility, stability and combinability of inscriptions are enhanced, how the networks are extended, how all the information is tied together in a cascade of rerepresentation” (Latour 1987: 246f). For social constructivists there is no way to model these overlapping processes of continuous composition and de-composition, of differentiation and fusion. What really happens cannot be objectified in modelling “objects” and the influence of society on them: “there is no pure object that first comes to the attention of the AI (Artificial Intelligence) manager […] and then is brought to the sociological manager for prioritising and dissemination. The sociological agency initiates and ‘runs’ the program” (Brannigan 1989: 606). The only software package which social studies of science would accept is the non-computational and implicit creation mode of society itself.

In the attempt to analyze the social construction of new knowledge within research groups, the theory of self-organization has proved to be a worthwhile approach. Originally used in natural sciences for the description of order arising out of disorder, it is nowadays also applied in human and social sciences for the analysis of complex issues.²

Simulations of complex social systems involve a challenge, especially in the selection of one or more programming techniques to model adequately parts of a real system. The theoretical guidelines in this context are given by Klüver (this volume); here, I wish to demonstrate some examples of simulations with a particular type of neural network, an interactive network, to show the kind of problems can be simulated with such a technique. Neural networks are only seldom used in the social sciences. One reason may be the fact that there are a lot of different types of neural network and it is often not clear — even for the expert — which kind should be used for a specific problem. In addition, the mathematical algorithms for neural nets are not easily understandable. Yet, despite these problems neural networks offer many interesting possibilities, especially for social scientists.

What are some criteria for modelling scientific creativity? What characteristics of creative scientific thought must be included to achieve historical veracity? How can cognitive science be brought to bear on these problems? This chapter suggests an approach and offers examples from which generalisations emerge.

Recent changes in the science and technology policies of a number of European countries are driven by the belief that a shift of academic activities towards more market-orientated research will increase national economic performance and prosperity. In the UK this has led to a move away from a predominantly peer review based funding system to one that explicitly incorporates industrial and political objectives. The central plank of the new policy regime is the UK Foresight Programme (previously Technology Foresight Programme). Its aims are two-fold. One objective is to identify opportunities where government involvement in R&D, and the science underpinning this, can enhance wealth creation. Another objective is to establish better communications and networks between science and industry — to affect the national innovation system — and thereby influence the behaviour of academic scientists through closer links. Other aspects of the reforms include the reorganisation of academic funding councils to include greater industrial representation, and the introduction of a ‘Research Assessment Exercise’ (RAE) employing various performance indicators to rank the research capabilities of different academic departments and institutions.