Edmund Chattoe

Sensitivity Analysis in the Social Sciences: Problems and Prospects

In the natural sciences and engineering, sensitivity analysis is a standard method for verifying simulation models. The term “sensitivity analysis” is generally used to describe a family of methods for altering the input values of the model in various ways and such analyses are included in the validation step of almost all technical simulations (Law and Kelton 1991, pp. 310ff). Nevertheless, a short review of simulation textbooks and other studies reveals that the term is currently used as a general catch all for diverse techniques: there is no precise definition and no special methodology currently associated with this term. A cursory inspection of recent work suggests that the situation is even worse in the social sciences.4 Most social simulators still omit any form of sensitivity analysis. (However, there are exceptions, mostly recent: Axtell et al 1996, Harrison and Carroll 1991, Latane 1996, Macy 1991, Saam 1996 and Saam and Reiter, submitted.) There is also a definite lack of methodological literature on sensitivity analysis in the social sciences.

TALKING ABOUT BUDGETS: TIME AND UNCERTAINTY IN HOUSEHOLD DECISION MAKING

This paper reports findings from twenty-six interviews conducted with retired households about the way they make budgeting decisions. The data are presented with three objectives. First, to describe the way that households carry out the process of budgeting. Secondly, to consider the implications for economic and sociological theories of consumption. Thirdly, to construct a framework for understanding budgetary decision making which organises the data usefully and places current theories in a clearer relationship to each other. It is found that personal budgeting has several characteristics which are poorly represented in existing theories of consumption: the strong relation between time and money in decision making, the importance of durable goods, the social rather than rational nature of household decision making and the need for budgeting strategies to deal with uncertainty. A five-level framework is presented which shows how different resources and methods are used by decision makers to manage their money.

A Simulation of Adaptation Mechanisms in Budgetary Decision Making

This paper describes a simulation which investigates the interaction and relative effectiveness of individualistic and social learning in producing workable budgeting strategies for agents with limited resources and bounded rationality. The simulation is motivated by an attempt to understand interview data collected from a sample of recently retired households about their budgeting behaviour. In particular, the simulation is designed to reflect, in simplified form, the information and decision processes typically available to individuals, given limitations of memory, accuracy and time for calculation. The paper also illustrates the inductive use of interview data as a basis for simulation, contrasting with the process of deductive modelling traditionally used for descriptions of economic activity. The simulation suggests that social learning is of considerable importance and leads to the emergence of differentiation in patterns of activity.

The Role of Agent-Based Modelling in Demographic Explanation

This chapter outlines five difficulties with modelling demographic behavior using Agent-Based Modelling (ABM), with particular reference to the role of social interactions in the diffusion of new contraceptive practices. The first difficulty is the potentially thin character of information and activities relevant to contraception within the wider framework of social action. The second difficulty is ensuring the correct relationship between ABM and existing theories, either based on aggregate statistical patterns or strong homogeneity assumptions about individuals. The third difficulty is ensuring an adequate representation of the contextual nature of social action, both in time and space. The fourth difficulty is to represent the complexity of decision processes adequately, so they can take account of the transmission of different kinds of information: norms, costs and benefits, practices and so on. The final difficulty is the possibility of collecting relevant data for building and testing ABMs. The chapter also draws attention to the connections between these difficulties and suggests solutions where these exist. In some cases, like the modelling of thin processes and cognitive complexity, considering social behavior from an ABM perspective reveals important challenges for social science research that are obscured by other approaches to modelling.

Building Empirically Plausible Multi-Agent Systems

Multi-Agent Systems (MAS) have great potential for explaining interactions among heterogeneous actors in complex environments: the primary task of social science. I shall argue that one factor hindering realisation of this potential is the neglect of systematic data use and appropriate data collection techniques. The discussion will centre on a concrete example: the properties of MAS to model innovation diffusion.

Developing the Selectionist Paradigm in Sociology

This paper clarifies and develops some of the arguments put forward by W.G. Runciman in his 1998 Sociology article ‘The Selectionist Paradigm and Its Implications for Sociology’. It intends to support his basic claim that mechanisms analogous to (but not synonymous with) natural selection are an important way of understanding both continuity and change in social systems. Nonetheless, it questions the emphasis of his discussion and extends his analysis of two substantive points. The argument proceeds in two stages. The paper begins by examining the many objections that Runciman rebuts and showing that many of them do not need rebuttal but are simply irrelevant to the selectionist paradigm. By irrelevant, I mean that the objections are logically flawed or simply do not apply to the selectionist paradigm as Runciman defines it. The purpose of this part of the paper is to sharpen the debate, so that attention can subsequently be focused on a smaller number of relevant objections that remain. The remainder of the paper attempts to open that debate by discussing the two relevant objections that appear most forceful. It attempts to show that, on closer examination, both objections are mistaken. The first objection is that human deliberation makes any analogy with random mutation in biology untenable. The paper argues that in fact selection and deliberation are complementary. Selection will continue to act on social practices to the extent that our models of the world are imperfect and our practices have unintended consequences. The second objection is that while selectionism is interesting, it may be irrelevant to sociological practice. The last part of the paper provides a more detailed analysis of an example used by Runciman, the task of explaining differing levels of male lethal violence across societies. This analysis suggests that while Runciman’s ‘discursive’ selectionist analysis (like functionalism) can generate suggestive hypotheses, appropriate techniques will be needed to transform those hypotheses into models which sociological research can ultimately test. It is argued that multi-agent computer simulation is a particularly suitable technique for representing evolutionary processes in social systems thus allowing selectionism to be put on the same sort of footing as other ‘middle range’ explanations like game theory and social network analysis.

Hunting the Unicorn

The chapter explores whether social simulation research has to adopt a realist epistemology, or whether it can operate within a social constructionist framework. Research on leadership in small groups is used to illustrate the argument. The main question to be explored is how one can understand the emergence of identifiable leaders from interaction in small groups. The focus is on the emergence of ‘macro’ phenomena (in this case, leadership and follower roles from an undifferentiated collection of agents) as a result of interactions between agents at the ‘micro’ level.