Special issue on “experimental economics and the social embedding of economic behaviour and cognition”
- First Online:
- Cite this article as:
- Heintz, C. & Bardsley, N. Mind Soc (2010) 9: 113. doi:10.1007/s11299-010-0082-1
- 142 Views
Can human social cognitive processes and social motives be grasped by the methods of experimental economics? Experimental studies of strategic cognition and social preferences contribute to our understanding of the social aspects of economic decisions making. Yet, papers in this issue argue that the social aspects of decision-making introduce several difficulties for interpreting the results of economic experiments. In particular, the laboratory is itself a social context, and in many respects a rather distinctive one, which raises questions of external validity.
Their social motives and social cognitive abilities
Their social environment
The main rationale for producing this thematic issue is that these social aspects of human decision-making are pervasive and important for economics. We also observe that current behavioural economics mainly relies, for its empirical data, on economic experiments. Whence the question: are laboratory experiments an adequate or sufficient method for investigating the social aspects of economic cognition and behaviour?
The papers in this issue reflect on the methods of experimental economics and the theorisation that it warrants, given that the social embedding of economic behaviour and cognition apply in the economists’ lab too. This gives rise to several questions. For example, can laboratory experiments really capture, or even take into account, the social aspects of human decision making? Are the effects of human social cognition and motivation really controlled in the lab? Is experimental economics a reliable method in the face of high human sensitivity to social factors when taking decisions? Experimental economists themselves have shown the context sensitivity of subjects’ behaviour; the label often applied is ‘framing effects.’ Notwithstanding this observed sensitivity, bringing subjects into the lab is often thought of as a means of isolating behaviour from the complexity of naturally-occurring contexts. The analyses of Bardsley, Baumard, Binmore and Guala suggest that, on the contrary, behaviour can sometimes be fundamentally altered by the experimental context in ways that are not monitored or understood by experimentalists. A central concern comes from the fact that economists’ laboratories and experimental settings are themselves distinctive social contexts. Among the problems that arise one could distinguish the following issues:
There are important differences between the social context of the experiment and the social context of people’s usual social environment. If such factors are changed, what does the behaviour in a laboratory context reveal about behaviour in a natural context? This is the problem of external validity, philosophical aspects of which are investigated in Guala (2005). It is explored in different ways in this issue by Baumard, Bardsley and Guala. There is a risk that the abstract and controlled environment of the lab, which enables the individuation of causal factors, has the unintended consequence of eliminating aspects of the environment that are centrally important to the economic phenomena of interest (see also Baumard et al. 2010).
The context of the laboratory is rather unusual, experientially, for the subjects: generally it does not resemble contexts of their daily life. The episode occurs in a laboratory, there are some often bizarre tasks to complete, and an experimentalist is instructing them about what to do (but supposedly leaving freedom to make choices). Given this oddity, how do subjects interpret the tasks? They will use their social cognitive abilities to work out what is expected of them, but their interpretation of the tasks may not fit the experimentalists’ own assumptions about how it is interpreted. This misfit can result from the pragmatics at work in interpreting the communicated instructions (Sperber et al. 1995) and it can be due to the fact that the subjects interpret the task with the help of their cultural knowledge. For instance, Binmore (this volume) mentions how some subjects interpreted a public good game as “the harambe”—a local economic practice. But the consequences of this may be ignored when the data are interpreted (see also Baumard 2010).
Furthermore, subjects may comply in the lab to norms that regulate social interactions in their usual cultural context; they may rely on routines that have been learned for solving problems that are specific to their usual social environment. The rationality of their behaviour can then be completely missed by the experimenter, who relies on his own understanding of the stakes of the experimental task. This point is explained and illustrated by Binmore’s paper (this volume; see also Heintz, 2005). More generally, psychological theories of ‘ecological rationality’ and ‘situated cognition’ suggest additional problems for the applicability of results obtained in lab environments to natural environments. Herbert Simon and Gerd Gigerenzer have insisted that human rationality is ecological: it is better understood as a match between cognitive processes and the environment in which they occur. Both Binmore and Baumard assert that the behaviour observed in games such as one shot public good games result from cognitive mechanisms or routines that are adaptive, even if they are not in the context of the lab. If so, the underlying cognitive processes for decision-making need not be interpreted as ‘maximizing utility’ when in the lab.
Economic experimental methods put subjects into new, artificial, environments. How can we understand the ecological rationality underlying agents’ decisions when the subjects are taken out of their usual environment? With economic cognition, knowledge of the usual decision setting should be highly relevant for understanding its ecological rationality. Thus, one way to fruitfully constrain interpretation and avoid misinterpreting agents’ behaviour in the lab is to appeal to other relevant theories and methods, including social anthropology and history, and evolutionary theory. For instance, Baumard and Guala (this issue) appeal to work in social anthropology and economic history in order to specify the relevance of decisions taken in the lab and how one might expect the results to generalise: these disciplines describe the natural environment in which decisions are normally taken.
Binmore (this issue) advocates using evolutionary game theory, together with economic and social history: the former specify which equilibria there are and how they can be attained, the latter recount how one equilibrium has come to be selected among the many possible ones. Binmore emphasises that equilibria are not necessarily found at once: cultural evolution and learning has to occur and it can take considerable time.
With time, however, economic agents may learn to take the best decisions for a given type of situation or problem. Bernasconi and Galizzi (this issue) analyse different learning rules that describe the learning dynamic observed in the lab, where subjects decide how to connect with others. The learning rules serve to analyse how subjects come to converge to an equilibrium … or not. In many situations, say Bernasconi and Galizzi, the chances of mis-coordination are very high. Their paper is an example where studies of evolutionary processes, models of cognitive processes (learning) and experiments are put together for studying economic systems and their dynamics.
We believe a strong case for interdisciplinary research in economics emerges from this issue. This is based on the fact that the data gathered by economic experiments provide an evidence base that underdetermines hypotheses about cognitive processes at work in subjects’ natural environment. Unfortunately, the interdisciplinary approach advocated above, making links with the (other) social sciences, is currently perhaps something of a fringe activity among experimental economists. A much more popular interdisciplinary approach is neuroecomics. But can neuroeconomics guide the interpretation of experimental economics? Bourgeois-Gironde (this issue) argues that neuroeconomics is not yet sufficiently advanced for guiding theorisation about the cognitive and affective processes at work in decision-making. His premise is that direct inferences from the activation of brain areas to the activation of cognitive processes or emotions are not warranted. Bourgeois-Gironde elaborates this critique with detailed analysis of several neuroeconomics studies, including work relevant to understanding the social motives behind economic decisions. He then specifies conditions for neuroeconomics to contribute reliably to the understanding of human decision-making. Although his paper concludes on an optimistic note on future prospects, the point is that neuroeconomics, in common with experimental economics, often relies on neo-classical theories and categories in interpreting the data. Is this continuing reliance healthy? There are diverse views on this question amongst experimentalists.
According to some,1 experimental economics is replacing the erroneous model of man that is the homo-economicus. A less radical view gives to experimental economics the task of refuting some of the assumptions of neo-classical economics, but leaving other aspects intact. Thus, the assumption that agents behave most of the time as if they were maximising their utility is still adhered to by many economists. For instance, Binmore (this issue) argues that in natural environment, people do find Nash equilibria.
Many experiments are conducted as a ‘critical friend’ to an economic model. In that perspective interpretation is, again, closely tied to theoretical models. This strategy is described and criticised by Bardsley (this issue), who argues that experiments which follow it often miss their target. Further, the model-centric view of experiments, Bardsley argues, rests on a denial either of the socially-constructed nature of the social world, or of this social construction. The article applies and extends arguments introduced in Bardsley et al. (2010), to consider the likely degree of external validity of model-implementing designs, particularly of an applied economics nature.
Last, but not least, economic experiments can be thought of and designed as means for revealing preferences. In that perspective, it is convenient simply to assume that subjects’ decisions actually reveal their preferences because they maximise their utility functions. Context-dependence tends to be abstracted away in the homo economicus model of modern microeconomics: only the objective function and factors determining the opportunity set are considered.
A very influential body of work done in this framework involves scholars such as Fehr, Gintis, Gätcher and Henrich. Their work is highly relevant to this thematic issue because it theorises about the social preferences people have on the basis of economic experiments.2 Such researchers do recognise, in a particular way, that economic behaviour and cognition are impregnated by social factors. Yet this approach comes under attack from several authors of this issue (Baumard, Binmore, Guala and, for neuroeconomic data, Bourgeois-Gironde). Some of its central assumptions are (a) that people have pro-social utility functions, including a preference for punishing free-riders, (b) that they maximize this function when making choices, be it in the lab or outside of the lab, and (c) that cooperation arises from this altruistic punishment behaviour. Binmore (this issue) is highly critical of this school of experimental economics, taking particular issue with assumptions (a) and (b). Baumard and Guala question the external validity of the experiments, and thus the interpretation given by the experimenters. For them, altruistic punishment as a behaviour observed in the lab is, contra Fehr et al., not at the origin of cooperative behaviour. Furthermore, it may not result from a significant preference for punishing free riders.
Bardsley opens up the debate by arguing that experiments are not shielded from external validity questions by sticking closely to theoretical models, and that sociality may actually be particularly problematic for certain model-implementing designs.
Binmore argues against the interpretation of experimental data in terms of ‘social preferences,’ in particular inequality aversion, on the grounds that behaviour has been removed from a context to which it may have been well-adapted, and that adaptation processes may take longer than the lab allows.
Guala responds to Binmore by arguing that competing explanations of laboratory data in terms of social preferences or evolutionary adaptation need to appeal to field evidence, rather than Binmore’s simulations, for a compelling resolution
Baumard reviews evidence, from anthropologists and economists’ work, that altruistic punishment as observed in the experimental settings, does not, in fact, play a role in the evolution of cooperation. He provides an alternative interpretation of experimental results based on an alternative model of the evolution of cooperation.
Bernasconi and Galizzi address social embedding by means of experimental economics, investigating the choices that people make when networking. They observe network formation in their experimental setting and theorise about the learning dynamics and subjects’ decision rules.
Bourgeois-Gironde applies the ‘reverse inference fallacy’ critique of neuropsychology to neuroeconomics, and also questions the ready interpretation of data in terms specified by economic modelling. He proposes that evolved neurological structures could be well adapted to modern settings and activities through a ‘recycling’ process and that the exploration of this hypothesis could offer a progressive way forward for neuroeconomics.
One could argue that the social aspects of economic decisions are, in the field of experimental economics, still striving to find the middle ground between the under-socialised agent ‘maximising utility’ and the over-socialised agent blindly abiding by norms. There is still much work to do since Granovetter (1985) pointed out a third way: considering economic agents as embedded in social structures. Behavioural economics could indeed clarify what this actually means in terms of the cognitive processes implemented by situated agents; but this also implies, as Granovetter already hinted at, a very good understanding of the social situations where these cognitive processes usually occur—their social embedding. We hope that the papers in this issue may help us to make progress in that direction.