Should System Dynamics be Described as a `Hard' or `Deterministic' Systems Approach?

INTRODUCTION

In this section the motivation and form of the paper are introduced. First, the main ideas of system dynamics are presented (readers familiar with the approach may wish to skip this subsection). The range of accusations that system dynamics is a `hard' or `deterministic' systems approach is then introduced. This provides the motivation for the remainder of the paper, the main sections of which are then outlined.

The Nature of System Dynamics

System dynamics is a system approach based on servo-mechanism theory (Richardson, 1991). It was created by Jay W. Forrester at MIT in the late 1950s and involves the modelling of social systems using computer simulation, with practitioners working closely with problem owners to structure debate about long-term policy. The history of the field is discussed elsewhere (Forrester, 1990; Lane, 1998) and its main ideas may be found in a range of publications (Forrester, 1958, 1961, 1968b, 1975; Randers, 1980a; Richardson and Pugh, 1981). In the account that follows, the field is presented in terms of three aspects: the ideas behind the practical modelling approach; the theoretical assumptions of the field; and finally its institutional features, its contribution to systems thinking and its literature.

As a modelling approach, system dynamics has three characteristics. First is the concept of information feedback loops. These involve the collection of information about the state of the system, followed by some influencing action which changes the state of the system. These closed loops of causal links involve delays and non-linearities as well as processes of accumulation and draining. The second characteristic is computer simulation. Although humans can conceptualize such loops, they lack the cognitive capability to deduce the consequent dynamic behaviour without assistance (Sterman, 1989, 1994). Computer simulation is therefore used rigorously to deduce the behavioural consequences over time of the hypothesized causal network. The shifting interplay of loops means that different parts of a system become dominant at different times. Such behaviour is counterintuitive, and may be explored using simulation models. The third and last characteristic of system dynamics is the need to engage with mental models. The most important information about social situations is only held as `mental models', not written down. These mental models are complex and subtle, involving hard, quantitative information and more subjective, or judgmental aspects of a given situation (Doyle and Ford, 1998; Lane, 1999a). Such models are the basis for organizational decision making. Hence, eliciting, debating and facilitating change in the mental models of decision makers can result in improved ways of managing a system. Modelling work must therefore be done in close proximity to problem owners, who are then able to see that their mental models are reflected in a computer model.

This means by which these ideas are implemented are shown in Figure 1. A situation is conceptualized in terms of loops and decision points. A system dynamics study is focused around a `dynamic hypothesis' -- the idea that a certain causal structure explains a certain dynamic behaviour. Model building tests this hypothesis using rigorous formulation and by synthesising objective and judgmental data (Randers, 1980b). Such a model is a theory for the structural source of the dynamic behaviour. When problem owners have confidence in a suitably validated model (Forrester and Senge, 1980), it can be used to test the dynamic hypothesis. It is by policy analysis and repeated experimentation with such models that organizational actors can learn together to create shared mental models. Subsequent policy making is therefore informed by an improved intuitive understanding of both the dynamic possibilities of the system and of the policy levers that might be used to steer it, and also to continue to learn about its operation.

[Figure 1 ILLUSTRATION OMITTED]

The theoretical assumptions of the field are more difficult to establish. Although a few publications have addressed this subject (Meadows, 1980; Barlas and Carpenter, 1990; Lane, 1994; Vennix, 1996), the field has tended to take its theoretical assumptions rather for granted. This concentration on trying to understand real-world phenomena perhaps reflects the field's solidly practical engineering roots (Richardson, 1991). A recent attempt to unearth the social theoretic foundation of system dynamics therefore had to rely primarily on inference from practice (Lane, 1999b). Various conclusions were proposed in that paper. On a superficial level, system dynamics appears to be locatable within the functional sociology paradigm of social theories, its ideas seeming to be a version of social systems theory (Burrell and Morgan, 1979). However, the craft of system dynamics, and hence its theory in use, has many links with more interactionist schools of thought and even some connections with interpretivism (see Lane and Oliva, 1998). This uncertainty leads to one possible conclusion that the theoretical position of system dynamics cannot be conveyed with reference to more traditional social theoretic assumptions but may best be seen in the light of those theories which seek to integrate the agency and structure views of the social realm (a suggestion explored elsewhere (Lane, 1999a, 1999c)).

Institutionally, system dynamics has created an International Society,(a) an annual conference and various national Chapters. There is a dedicated refereed journal, System Dynamics Review. The field has produced numerous publications in book and article form which record its achievements in theory building. Comprehensive lists may be found elsewhere,(b) so for the purposes of this paper it is only necessary to mention a sample. System dynamics has been applied to: the growth and stagnation of cities (Forrester, 1969), global development (Forester, 1971b; Meadows et al., 1972, Meadows et al., 1992), software development (Abdel-Hamid and Madnick, 1990), emerging economies (Saeed, 1991), health care policy (Lane et al., 1999), organisational learning (Senge, 1990) and the implementation of TQM initiatives (Sterman et al., 1997). A very useful collection of theoretical and practical articles is available in Richardson (1996).

On the `Hard' or `Deterministic' Criticisms of System Dynamics

Amongst the systems scientists who have offered comments on it (and the few social scientists who have come across it), system dynamics is frequently judged to be a `hard' systems approach. This does not mean `difficult'. Rather, it likens the approach to systems engineering and to social systems theories which conceptualize the social world as operating like a machine -- both of which views modern systems science has felt a need to escape from. In fact, in the struggle to find criticisms to hurl, a barrage of terms has been employed. In addition to `hard' (Keys, 1988; Dash, 1994), system dynamics has also been labelled as `simple' and `unitary' (Jackson and Keys, 1984), `machine'-like (Flood and Jackson, 1991) and `deterministic' (Jackson, 1991, 1994). Many of these terms overlap. For example, having asserted that system dynamics takes a deterministic view of human behaviour, Jackson (1991) proceeds to offer this as the reason why `Forrester's modelling techniques have tended to be used in conjunction with essentially hard systems methodologies' (Jackson, 1991, p. 93).

These criticisms can be interpreted in at least four different ways (see below). For the purpose of this paper the term `hard/deterministic' will therefore be used to describe this bundle of criticisms.

Why are these criticisms important? As already mentioned, they make a highly unwelcome linkage back to systems approaches now discredited as examples of `hyper-rationalism' (Rosenhead, 1989; Lane, 1994). They have also led to the conclusion that the field is `caught in an appalling paradox' (Jackson, 1993, p. 22). Further examples are given in the following sections.

This is strong stuff. Such criticism would be an important verdict for any research program. What is rather more galling is that at least two of these criticisms are, as demonstrated below, false. Yet these criticisms are widely known within the general systems science movement. For members of that movement, the above citations are possibly the most important information sources on system dynamics (notwithstanding the fact that all originate from researchers outside the system dynamics field). The systems science movement is diffusing into the social sciences via information systems, planning theory, operational research, management etc. This is creditable but it is happening with other systems approaches, system dynamics seeming to have fallen into disrepute in the eyes of some, partially because of its imputed hard/ deterministic nature. For system dynamics to fail to prosper because of genuine weaknesses and inconsistencies is one thing. For such failure to occur because of incorrect readings of the field's core ideas is an entirely different matter.

To begin to counter this situation the following sections deal with the various `hard/ deterministic' criticisms that are made of system dynamics. These can be interpreted in four different -- though not entirely distinct -- ways. These are outlined below and illustrated in Figure 2:

* The first accuses system dynamics of philosophical naivete in assuming that future events can be prophesied.

* The second accuses system dynamics of dehumanizing extremism in assuming complete structural control of the decisions of human agents.

* The third accuses system dynamics of social scientific crudity in assuming that there are cause and effect laws which exist outside of human subjectivity in a form of timeless `grand theory'.

* The fourth and last accuses system dynamics of being operationally austere and coercive in that it is essentially a `systems engineering' approach.

[Figure 2 ILLUSTRATION OMITTED]

For ease of exposition this paper is structured around these four interpretations of the term, although it is acknowledged that the ideas, are related and somewhat overlapping. In each case the criticism is described in more detail and a response is then offered. By the close of this paper it will have been argued that a majority of the criticisms are unfounded, while the terms of debate regarding the remaining criticisms have been clarified constructively.

PROPHECY CREATION?

The first interpretation of `hard/deterministic' is that it indicates the philosophical position that `all things, including will, have causes' (Beishon and Peters, 1981, p. 323). In this view a clockwork universe rolls forward in time in an utterly mechanical way, with all phenomena -- events, thoughts and actions- as mere inevitable consequences, rationally predictable in principle. One can then obtain a `prophecy', a `prediction about an event which we can do nothing to prevent' (Popper, 1957, p. 43). What is then called for is not a proactive response but a …