System dynamics, as introduced by Jay Forrester during the late 1960s (Forrester, 1968), is a modelling paradigm which allows for the representation of the dynamics of systems’ behaviours in terms of variables, processes and mechanisms. Forrester was a pioneer in the computational and simulation modelling of social systems: he employed system dynamics to model the dynamics of change processes in large social structures. Taking advantage of the rise of computer technology and advances in computer simulation, Forrester (1961, 1969, 1973) employed it in forecasting the behaviour of both industrial and social systems.
His works fundamentally changed the modelling perspective over social and political modelling. Moreover, his views challenged experimental psychology as well as social and cognitive psychology to use the system dynamics paradigm for the modelling of behaviours, beliefs and attitudes. For almost two decades during the 1950s until the late 1960s, attitude modelling research in social and political psychology was dominated by behaviourist modelling. The system dynamics paradigm provided perhaps the most powerful reinforcement to the idea that behaviours can be controlled from outside by messages with a certain frequency and conceptual configuration. However, system dynamics offered much more than this, competing with the behaviourist paradigm until weakening its position in modelling research: it provided a way to model the behaviours of the system by taking into consideration both internal and external stimuli as well as the system structure. This attracted the attention of social and political psychologists, who found this view appropriate for attitude modelling. The system dynamics view on behaviour and attitude change was influential in the first computer simulation model developed by McPhee and collaborators in the late 1950s and the beginning of the 1960s for modelling aggregate tendencies in the voting choice of the electorate in presidential elections.
The idea of combining system dynamics and computer simulation was reinforced by the contribution of another area of intensive research during the 1950s, namely the propaganda studies and persuasive communication research developed by Carl Hovland and his team at Yale University. One of the most relevant studies during the 1950s is the Yale model (Hovland et al., 1953), which assumes that, in political propaganda, persuasion is achieved by the ‘source–message–receiver’ scheme, in which both the message and some attributes of the source, like its credibility, authority and competence, are essential for the quality and effectiveness of persuasive communication. The influence of this model in political attitude research has been tremendous: at that time, no modelling approach would have escaped it. On the contrary, both social psychology and political psychology have intensively employed it in explaining political attitude formation and change. Its influence diminished only after cognitive theories had substantially penetrated social and political psychology research.
One last detail would appropriately complement this already complicated picture: computational modelling and simulation at the time was in a period of rapid growth, and strong enough and quite promising so as to attract the eyes of electoral strategists for electoral campaign and electoral media communication designs. Pioneered by William McPhee, computer simulations seemed very attractive, though unfortunately extremely expensive. And the most prohibitive costs were not those for acquiring computer machines, but those for computer programming and modelling expertise. The history of the rise and growth of experimental political science would tell much about the times of this painful experience. Computers and computer simulation advances were in constant development starting in the 1950s. System dynamics was the first computer simulation paradigm and has been employed in political decision‐making, electoral studies and political behaviour research.
System dynamics was originally designed for modelling economic processes, in which the dynamic behaviour of a variable is studied by computer simulations. A computer simulation was used as a means to implement the model by varying both internal (state) and external variables (stimuli).
The behaviourist paradigm models a system as a black box and a system’s behaviour as solely determined by external stimuli, which could either reinforce or punish certain types of behavioural variations. The main characteristic which made the systems dynamics paradigm so appealing for social and political psychology research concerned its ability to model the behavioural change: the system’s behaviour is viewed as influenced by both internal processes (attitude change, belief change) and external stimuli (messages) such that the model includes one or multiple feedback loops which contribute to the causal explanation of the behavioural change.
In systems theory, a system is characterized by a dynamic state. It could vary as a function of factors defining both internal and contingent sources of variation. The dynamics of a system are described by a rule which always identifies the next state following from the current state under certain classes of external and internal stimuli. The system dynamics paradigm provides for a causal view upon the structure and behaviour of a system. As a difference from the behaviourist paradigm, which achieves a system’s behaviour in open‐loop representation, the system dynamics paradigm achieves a system behaviour in a closed‐loop representation. Behaviourist modelling is based exclusively on external control of the behaviour, whereas in system dynamics the system’s behaviour is modelled as a dynamic feedback loop in which both external and internal structures are involved (Levine, 1983: 326–327).
A system dynamics model is represented as a causal loop, which is often described as a ‘stock and flow’ model: a ‘stock’ is always a variable whose quantitative value could dynamically increase or decrease, while a ‘flow’ is a rate of change process, which could also dynamically increase or decrease. Their variations are described by differential equations which represent the rules describing the state change. The system dynamics model could be simulated on a computer such that the dynamic transfer from one state to another (i.e. the system’s trajectory in time) could be registered and further studied with quantitative analysis methods.
In individual and social psychology, as well as in political psychology, the attitudes and beliefs are the variables which cover relevant aspects of internal sources of change.
The Hovland–Yale model of persuasive communication inspired a ‘source–message–receiver’ modelling of attitude and belief change. Attitudes and beliefs are subjected to persuasion processes in which messages with a particular conceptual configuration are received from external sources, either individuals or groups or media campaign communications. Approached in the system dynamics paradigm, attitude and belief change has been modelled as a dynamic variation and described with the help of differential equations.
The system dynamics approach to attitude and belief change modelling as well as the inspirations coming from other disciplines, like communication, cognitive psychology or information processing, has succeeded in shaping a class of modelling theories able to answer the demands and expectations formulated in social and political psychology research with concern to several relevant issues: decision‐making and preference aggregation in mass publics. This class of modelling theories has opened the door for the development of computational and simulation modelling of political attitudes. Though not actually a movement, this research area comprised famous scholars and groups known by their topics of interest and the solutions they have provided to major research questions.
One of these groups was conducted by John E. Hunter, who laid the basis of system dynamics modelling of attitude change. John E. Hunter and collaborators founded a basic trend in attitude change research as they developed a class of mathematical models for simulating the dynamics of attitude change in the most important paradigms: behavioural, information processing, cognitive dissonance (Levine, 2003). In these models, attitude change is modelled as an effect of persuasive communication, and influenced by both message and source.
Hunter’s work was further developed in two main (otherwise, closely tied) research areas of attitude and belief change modelling: (i) spatial‐linkage models, which include several classes of multidimensional attitude–belief change models, like variable mass models, spring models, discontinuous models and hierarchical models; (ii) multidimensional ideology formation and change models.
The class of attitude change models based on the system dynamics paradigm further inspired Joseph Woelfel and his collaborators to extend Hunter’s issue of dynamic attitude change to the study of spatial modelling of attitude and belief change. They developed the Galileo Model, a theory which provides support to several classes of spatial‐linkage models. These classes of models address the problem of measurement of attitude change in multidimensional cognitive spaces. Though actually a measurement model, Galileo is based on a cognitive theory about attitude formation and change inspired by the early research on the semantic space (Osgood and Tannenbaum, 1955), balance theory (Heider, 1946) and cognitive dissonance theory (Festinger, 1957).
Some of the research issues initially developed by John Hunter and collaborators, like the hierarchical models of attitude–belief change, were further developed by Edward Fink and his team: hierarchical models of inter‐attitudinal change address the issues of dynamically changing cognitive systems and structures (Dinauer and Fink, 2005).
Also initiated by Hunter, the issue of ideology change was further developed by George Barnett, Kim Serota and others into modelling approaches to ideology as both influencing and being influenced by political communication in large masses (Barnett, 1974; Serota et al., 1975).