The science of uncertainty II

In the last post we introduced the work of Peter Allen, who models complex phenomena using non-linear computer programmes. These models do not assume that each of the interacting agents is the same, nor do they assume that the interactions happen at an average rate. The parameters set for the model do not oblige it to move towards equilibrium. With these programmed assumptions Allen describes the way in which interacting agents adapt to and learn from each other, and how novelty arises from deviant and eccentric behaviour. Allen concludes that ‘strategy’, which we might take to  mean a description of what all the agents are doing, develops imperfectly and can only be made sense of retrospectively.

In this post we will explore the ideas behind agent-based modelling further as a way of pointing out some of the similarities between scholars operating within different disciplines but coming to some similar conclusions about uncertainty and complexity.

Another researcher Peter Hedström, of Oxford University, describes himself as an analytical sociologist, and also develops computer models to explain complex social events. He has, by his own definition, very little patience with theoretical sociology because it seems to him too ‘imprecise’. He metes out particular criticism for one of the sociologists I refer to a lot, Pierre Bourdieu, for offering explanations of social phenomena that mystify as much as they explain. they are for him insufficiently scientific. Instead he has tried to explain social phenomena by simulating them mathematically like Peter Allen, because he regards this as a precise, abstract and realistic method.

Hedström carefully sets out his position in trying to explain a theory of action that is appropriate for the purposes he has in mind. He rejects both deterministic sociological theories and ones derived from classical economic theories of action which imply an atomised and heroic actor. The former theories, her argues sometimes portray actors as passive subjects whose behaviour is explained by causal factors of which they are usually unaware. Meanwhile economic theories classically portray an actor equipped with unlimited cognitive abilities to choose consistently the optimal course of action. His models understand the intentional actor to be embedded in concrete, ongoing social relations.

The model of action that he is developing is intended to describe the mechanisms of a sociological theory, but not as a predictive tool, nor as a method of interpretation for understanding a concretely existing actor. This involves his developing ideal-typical actors operating according to typical intentions that could motivate typical actors to behave in specific ways. Hedström describes his models as analytical because they remove any factors which are deemed to inessential for studying the problem at hand, and he considers them realist because the elements which are retained are thought to reflect the real processes at work.

It is already apparent that Hedström’s models rely upon quite strong assumptions of consciously intentional behaviour expressed by autonomous ‘typical’ actors operating devoid of other factors deemed irrelevant. They act reasonably rather than rationally embedded in interactions with others. In order to programme his agents with intentions he assumes that they operate with a varied combination of desires, beliefs and opportunities (DBO) which affects their interactions with local others. The assumptions gives him a certain degree of queasiness as a positivist scientist since mental states are not observable, but drawing on other scholars he continues with them because of the simplicity and clarity that they provide. The model is based on assumptions which are thought to be roughly correct in the real world in which agents operate.

He argues against the tendency of some sociologists to separate out social reality as though it occurred on different stratified levels. He is also critical of quantitative social research methods which use survey data: he argues that they produce rich data on the attributes of individuals but no data on the actions with whom these people interact. The surveys are based on the assumption that the whole can be understood by studying individuals in isolation from each other. For Hedström, social phenomena are emergent and the social and the individual mutually and dynamically influence each other. By emergence Hedström means that social properties cannot be predicted in advance even by knowing everything there is to know about the pre-emergent properties of the parts. Knowing the behaviour of the isolated parts leaves us a long way from knowing the whole and the social processes which have contributed to it.

From various empirically calibrated agent (ECA) computer models he has developed, including one which attempts to explain patterns of unemployment and job seeking among young people in different districts of metropolitan Stockholm, Hedström draws some interesting conclusions:

1                    There is no necessary proportionality between the size of a cause and the size of its effect.

2                    The structure of the social interaction is of considerable importance in its own right for the social outcomes that emerge.

3                    The effect a given action has on the social can be highly contingent upon the structural configuration in which the actor is embedded.

4                    Aggregate patterns say very little about the micro-level processes that brought them about.

It is difficult to disaggregate these observations since Hedström intends them to be taken together, but in analysing the interaction of agents within a network he concludes that small variables can make a big difference to outcomes. While in one situation the actions of X might lead to Y, in another context where the power relationship between X and the network of agents they are related to are slightly different, an entirely different outcome is possible even if the same actions are pursued. Moreover, Hedström also admits that other factors, outside the field of scrutiny, can also have a big impact on outcomes between interactions. Social patterning arises in unpredictable ways even if we can identify many of the important factors, and in addition to this there are other factors that we cannot identify which may also influence the outcome. So rather than being linear and predictable, he concludes that the relationship between the individual and the social, the individual agent and multiple agents, is ‘complex and precarious’ where ‘large scale social phenomena that are observed may simply be due to an uncommon combination of common events and circumstances’.

Hedström is open to the critique that modelling complex social phenomena in a laboratory might be seen as merely contingent to mainstream sociological research which studies real human beings. What he thinks it offers, however, is a more realistic theoretical insight into dynamic social processes.

The reason for bringing in Hedström and Allen is as a way of pointing to social researchers who work within a natural science tradition, but who nonetheless claim that social phenomena are non-linear, inherently unpredictable and arise from micro-interaction between agents acting locally. Self-organising agents in asymmetric relationships produce global patterns which in turn constrain local interaction.

In the next post I will write about the work of more theoretical social scientists who do not aspire to using the formalisation of mathematics, but nonetheless seem to describe non-linearity and unpredictability in similar terms. In doing so I will be trying to reconcile the insights of selected natural and social scientists working in the domain of uncertainty.


One thought on “The science of uncertainty II

  1. Rob Warwick

    I do think it is fascinating that people look to complex mathematical formulae to explain everyday life and then seek to apply this understanding in areas that we can all relate to, ie economic models of the Scottish fishing industry in the case of Allen. Perhaps I am being unkind; his work does offer useful insights. Whereas the likes Bourdieu and DeCerteau seem to go for the jugular of what we do in our social lives. Having tinkered around economic and computer models I am always interested in how they are taken up and the type of conversation they lead to. In this sense, the computer stuff is only part of the story (it is, in Stacey’s words, the abstraction). All those graphs, complicated differential equations, statistical formulae that I distantly remember from my science days seem to have a soothing and mesmerising effect (… look into my eyes, … the eyes, not around the eyes, don’t look around my eyes, look into my eyes, you’re under – Kenny, Little Brittain). They seem to allow people to make decisions, bolster their confidence that “it” will work because “it” is rational; rational because it is predictable. All of this without thinking what it is we are doing when we come together to make sense of things and make decisions, which takes us back to Bourdieu and DeCerteau.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s