Can cognitive science create a cognitive economics?
Nick Chater
Behavioural Science Group
Warwick Business School
[email protected]
Cognitive science can intersect with economics in at least three productive ways: by providing richer models of individual behaviour for use in economic analysis; by drawing from economic theory in order to model distributed cognition; and jointly to create more powerful 'rational' models of cognitive processes and social interaction. There is the prospect of moving from behavioural economics to a genuinely cognitive economics.
N. C. was supported by ERC grant 295917-RATIONALITY, the ESRC Network for Integrated Behavioural Science [grant number ES/K002201/1], the Leverhulme Trust [grant number RP2012-V-022]. Research Councils UK Grant EP/K039830/1, and the Templeton Foundation.
Cognitive science views thought as computation, and aims to construct theories, and build models, of the computational processes "inside the head." Economics, by contrast, has traditionally avoided peering at our cognitive innards. It has, instead, treated individuals "as if" they behave as fully rational agents, and use strong assumptions of rationality to make analytically tractable calculations about how such agents will behave in aggregate, for example, in markets.
How might these two disciplines be brought closer together? What might a "cognitive economics" look like? Here, I wish to explore three promising, and very different, links. First, and perhaps most obviously, the "rational economic agent" may be enriched by understanding the operation and limitations of the cognitive mechanisms from which economic and other decisions are constructed. Thus, cognitive science can export knowledge into economic realm. Second, economics provides powerful technical machinery for analysing distributed information processing systems: markets, not just minds, can be thought of as computational machines. Cognitive science may benefit from importing some of this machinery, to understand distributed cognition across people, and perhaps also information processing within the person. Third, and perhaps most promising, cognitive science and economics can jointly contribute to new rational theories of human thought and behaviour, upon which accounts of individual and collective behaviour can be constructed. Here, I briefly illustrate, in turn, the opportunities that may be available if we further explore each of these links, before concluding.
Exporting cognitive science to economics
One the founders of cognitive science, Herbert Simon, was, at least in part, motivated to understand the machinery of individual problem solving and decision making by the belief that the increasingly stringent assumptions of individual rationality embodied in late twentieth century mathematical economics may "idealise away" many important economic phenomena. Simon (1956; see also Cyert & March, 1963) argued that the information problems faced by firms, individuals, and, indeed, specific cognitive mechanisms within individuals, are spectacularly complex, so that the attempt to optimise is hopeless. Instead, the brain must, for many problems, be content to satisfice---i.e., settle on 'good enough' solutions, using heuristic strategies, rather than optimisation.
The last 40 years has seen explosive growth in the study of non-optimal judgements and decisions (e.g., Kahneman & Tversky, 2000) and attempt to build links to real-world economic phenomena, such as stock market booms and crashes (De Martino, O'Doherty, Ray, Bossaerts & Camerer, 2013). The subfields of behavioural economics and behavioural finance have, as their names imply, typically focused on behavioural anomalies and their implications, rather than exploring the impact of underlying cognitive mechanisms. A deeper integration of the study of individual behaviour and the operation of economic institutions such as the market may require exploring underlying cognitive foundations, a process that has already begun. For example, theories of learning based on reinforcement and heuristic decision making rules have been widely applied in economic contexts (e.g., Gigerenzer & Selten, 2002). Moreover, general principles of cognition, such as the apparent comparative nature of perceptual judgement (so that subjective magnitudes cannot be encoded in absolute terms), may directly undermine the idea that people can have stable valuations for 'pleasures,' 'pains,' and perhaps more abstract concepts such as risk and delay (Stewart, Chater & Brown, 2006; Vlaev, Seymour, Dolan & Chater, 2009). Similarly, the apparent 'seriality' of some aspects of cognition (e.g., Pashler, 1999) may impose restrictions to one-reason decision making (Gigerenzer & Goldstein, 1996), and revise standard models of multiattribute choice which integrate many pieces of information simultaneously in favour of serial process models (e.g., Busemeyer & Townsend, 1993; Tsetsos, Usher & Chater, 2010; Tversky, 1972). Equally, though, cognitive models based on massive parallel information integration (e.g., Kunda & Thagard, 1996) may equally provide a basis for non-standard economic analysis. A particularly interesting question is where creating richer cognitive models into economics introduces complexity that is merely "averaged out" when many agents operate in a market context (e.g., because agents have incentives to exploit, and therefore to reduce, any market anomalies arising from behavioural biases); and where, by contrast, richer cognitive and behavioural models can provide qualitatively new predictions concerning economic behaviour (Frey & Eichenberger, 1994).
Importing economics into cognitive science
Economics is founded on rational choice theory; yet so, increasingly, is cognitive science. Marr (1982) argued that the starting point for cognitive scientific explanation in any domain is a "computational level" theory which clarifies the nature of, and solution for, the computational problem that the brain is addressing. Only then can algorithms be proposed which may instantiate, or at least approximate, that solution; and, finally, neural mechanisms that might instantiate such algorithms can be hypothesised.
At Marr's computational level, problems of decision-making and action are naturally modelled in terms of maximising expected utility. For example, we may suppose that the motor system has some objective (e.g., successfully catching a ball, picking up a cup) balanced against 'penalties' for inaccuracy, excessive energy costs, unnecessarily jerky trajectories and so on. A rich theory of motor control based on Bayesian decision theory has been developed and has been experimentally well-confirmed (Koerding & Wolpert, 2004; Wolpert & Landy, 2012). Similarly, also at the computational level, problems of perception, learning, and reasoning are naturally modelled as Bayesian inference: prior probabilities of hypotheses about the world are updated, conditional on the arrival of new data, to produce posterior probabilities in accordance with Bayes' rule, an approach that has been widely applied (Anderson, 1990; Oaksford & Chater, 2007; Tenenbaum, Kemp, Griffiths & Goodman, 2011).
Bayesian decision theory and Bayesian updating are part of the common inheritance of both cognitive science and economics, deriving from mathematics and statistics. But rational choice theory, and its variants, has developed in economics in ways that have yet to be exploited in cognitive science.
One area in which economists may have important contributions to make to rational models of cognition comes from non-standard formal approaches to probability and decision theory (Machina, 1990), which may be required to model how agents deal not merely with risk (where the probabilities of all possible outcomes are known) with full-blown uncertainty. A range of models to deal with uncertainty, and our aversion to it, have been proposed in economics, and some have been applied to explaining economic phenomena (e.g., returns on risky and risk-free assets across the business cycle, Collard, Mukerji, Sheppard & Tallon, 2011). Given that almost all real-world behaviour involves interacting with a world which is little understood, and hence a matter of uncertainty rather than sharply defined risk, we might suspect that such models may also be valuable for cognitive science. Moreover, extensions of rational models, for example, to deal with scenarios in which a cognitive agent has to enlarge its hypothesis space (e.g., in the light of scientific discovery technological advance) are likely to be applicable in cognitive science (Karni & Vierø, 2013). Standard Bayesian models, rather implausibly, require that priors are assigned to all hypotheses that will ever be entertained, at the outset.
The operation of the market has been viewed as a computational problem: to find the most efficient allocation of goods and services (Axtell, 2005). This problem is most naturally solved in a distributed fashion, by profit maximising agents, whose links with behaviour of other agents is primarily mediated by the price mechanism—this is an extension of Adam Smith's (1776) concept of the 'invisible hand,' which, under some circumstances, guides the self-interested butcher and baker to carry out actions which, as it turns out, contribute to the common good. This suggests the possibility that, if the brain is itself a radically distributed system, it may use similar mechanisms to coordinate its activity (for some suggestive computational explorations, see, Baum, 1999).
Linking with the discussion in the previous section, it is natural to wonder how well the 'invisible hand' work when the agents are not the hyper-rational agents of conventional economics, but instead are the limited, capricious,
A joint project: cognitively informed rational models
While the import and export of ideas between the cognitive science and economics is likely to be productive, there are also many important cognitive domains in which rational theories are underdeveloped, and were both cognitive scientists and economists are likely to have important contributions to make.
One of the deepest puzzles in the cognitive and social sciences concerns the rational response to inconsistent beliefs or preferences. In the context of economics, if market prices are inconsistent, then a clever trader or entrepreneur can make money for sure by arbitrage (buying at the low price, and selling at a high price), and the very act of arbitrage will narrow the price inconsistency. But, in general, inconsistent prices can be made consistent in indefinitely many ways. Similarly, in cognitive science, if beliefs or preferences are inconsistent, there is the same infinity of ways to 'fix' the problem—normative theories, such as logic, Bayesian inference, and decision theory can reveal the inconsistency--- but make no recommendation as to how to solve it. One reason why the economic version of the problem may prove to be more approachable is that, while beliefs and preferences within a single individual are not easily observed (where, indeed, they are well-defined at all), the trajectories of market prices are often fully observable, as arbitrage opportunities are revealed and closed.
Another deep, but perhaps more tractable, challenge for both disciplines is to model social behaviour. Economics has increasingly favoured game theory, and the notion of the Nash equilibrium, as a foundation for interactions: these ideas have provided a rigorous foundation for many aspects of microeconomics.
Yet the conventional Nash equilibrium concept may be too individualistic to model many aspects of social behaviour, in which people engage in joint actions (Bratman, 2014; Sebanz, Bekkering & Knoblich, 2006). When we are engaged in joint action, it appears crucial that each of us can reason about what 'we' are attempting to do (e.g., lift the table); inferring the best (in some sense) approach to solving the problem (I stand at one end, you at the other); and hence inferring which role each of us should play. Empirical research is increasingly revealing the ubiquity of joint action, and cognate notions such as joint attention (e.g., Moore & Dunham, 2014); and joint action and attention may, indeed, be fundamental to communication and language (Tomasello, 2008). But joint action is also puzzling for theories of how we understand other minds in cognitive science: such accounts, whether based on folk theories (Wellman, 1990) or simulation (Gordon, 1986), typically propose that each person is attempting to second-guess what the other will do, which can lead to regress. Non-standard versions of game theory, such as team reasoning (Bacharach, 2006; Sugden, 2003) and virtual bargaining (Misyak, Melkonyan, Zeitoun & Chater, 2014) may be the starting point for more adequate accounts.
A final challenge for both cognitive science and economics is to understand, from a rational standpoint, the cognitive value of language and communication. If we view communication as action, and assuming that the communicator is choosing its action to maximise its expected utility, rather than communicating what is, for example, true, then it is unclear why communication can be anything other than 'cheap talk' (Farrell & Rabin, 1996). The economic theory of signalling can explain how some actions can, nonetheless, have communicative value, when difficult to 'fake' (e.g., a job seeker might signal their 'quality' to a potential employer by passing difficult exams, Spence, 1973). But this type of account is inapplicable to normal communicative interactions, where talk is indeed cheap. Understanding how communicative acts can be both acts, and can have communicative significance remains deeply puzzling from the point of view of a rational theory of action. Economists have made some progress in addressing these challenges (e.g., Glazer & Rubinstein, 2006; Rubinstein, 1998; as have philosophers, e.g., Skyrms, 2010). In particular, perhaps the communicative significance of a communicative signal (e.g., an utterance, gesture or facial express) comes from solving a 'coordination problem' (Schelling, 1960) between sender and receiver: communication succeeds just when both parties 'coordinate' on the same interpretation of the signal. This viewpoint suggests that communication is a type of joint action (Tomasello, 2008), as discussed above; and that a rational theory for joint action may therefore be a foundation for a rational theory of communication, may, perhaps, be a promising starting point.
Towards a cognitive economics
Cognitive science focuses on computational problems within the individual; economics can be viewed as focusing on computational problems which are solved by aggregations of individuals organised into institutions, such as markets and firms. Economics may profit substantially from mechanistic constraints and insights from cognitive science; rational models of cognition in cognitive science mean benefits essentially from importing powerful methods from economics. And the insights and methods from both disciplines may be required to deal with some of the deepest questions in the cognitive and social sciences, such as how do we resolve inconsistency, what underpins joint action, and understanding the communicative value of language.
References
Axtell, R. (2005). The Complexity of Exchange. The Economic Journal, 115, F193-F210.
Bacharach, M. (2006). Beyond individual choice: teams and frames in game theory. Princeton, NJ: Princeton University Press.
Baum, Eric B. "Toward a model of intelligence as an economy of agents." Machine Learning, 35, (1999): 155-185.
Bratman, M. E. (2014). Shared agency: A planning theory of acting together. Oxford: Oxford University Press.
Busemeyer, J. R., & Townsend, J. T. (1993) Decision Field Theory: A dynamic cognition approach to decision making. Psychological Review, 100, 432-459.
Collard, F., Mukerji, S., Sheppard, K., & Tallon, J. M. (2011). Ambiguity and the historical equity premium. Department of Economics Discussion Paper Series (Ref: 550 ), University of Oxford.
Cyert, R. M., & March, J. G. (1963). A behavioral theory of the firm. Englewood Cliffs, NJ: Prentice Hall.
De Martino, B., O'Doherty, J. P., Ray, D., Bossaerts, P. & Camerer, C. F. (2013). In the mind of the market: theory of mind biases value computation during financial bubbles. Neuron, 79, 1222-1231
Frank, R. H. (1988). Passions within reason: The strategic role of the emotions. New York: WW Norton & Co.
Frey, B.S., & Eichenberger, R. (1994). Economic incentives transform psychological anomalies. Journal of Economic Behavior and Organization, 23, 215–234.
Gigerenzer, G. & Goldstein D., G. (1996). Reasoning the fast and frugal way: Models of bounded rationality. Psychological Review, 103, 650–669.
Gigerenzer, G., & Selten, R. (Eds.). (2002). Bounded rationality: The adaptive toolbox. Cambridge, MA: MIT Press.
Glazer, J., & Rubinstein, A. (2006). A study in the pragmatics of persuasion: a game theoretical approach. Theoretical Economics, 1, 395-410.
Gordon, R. (1986). Folk Psychology as Simulation. Mind and Language, 1, 158–71.
Kahneman & A. Tversky (Eds.) (2000). Choices, values and frames. New York: Cambridge University Press and the Russell Sage Foundation.
Karni, E., & Vierø, M. L. (2013). "Reverse Bayesianism": A choice-based theory of growing awareness. The American Economic Review, 103(7), 2790-2810.
Koerding, K. P. & Wolpert, D. M. (2004). Bayesian integration in sensorimotor learning. Nature, 427, 244–247.
Kunda, Z. & Thagard, P. (1996). Forming impressions from stereotypes, traits, and behaviors: A parallel-constraint-satisfaction theory. Psychological Review,103, 284.
Machina, M. J. (1990). Non-expected utility hypothesis. The New Palgrave: Utility and Probability, pp 74-84.
Marr, D. (1982). Vision. San Francisco, CA: Freeman.
Misyak, J., Melkonyan, T., Zeitoun, H., & Chater, N. (2014). Unwritten rules: Virtual bargaining underpins social interaction, culture and society. Trends in Cogntive Sciences, 18, 512-519.
Moore, C., & Dunham, P. (Eds.). (2014). Joint attention: Its origins and role in development. Hove: Psychology Press.
Oaksford, M., & Chater, N. (2007). Bayesian rationality the probabilistic approach to human reasoning. Oxford, UK: Oxford University Press.
Pashler, H. E. (1999). The psychology of attention. Cambridge, MA: MIT press.
Farrell, J., & Rabin, M. (1996). Cheap talk. The Journal of Economic Perspectives, 10, 103-118.
Rubinstein, A. (1998). Economics and language. Foerder Institute for Economic Research and Sackler Institute for Economic Studies, Eitan Berglas School of Economics, Tel-Aviv University.
Schelling, T. C. (1960). The strategy of conflict. Cambridge: Harvard University Press.
Sebanz, N., Bekkering, H., & Knoblich, G. (2006). Joint action: bodies and minds moving together. Trends in Cognitive Sciences, 10, 70-76.
Skyrms, B. (2010). Signals: Evolution, learning, and information. Oxford: Oxford University Press.
Simon, H. A. (1956). Rational Choice and the Structure of the Environment. Psychological Review, 63, 129–138.
Smith, A. (1776). An Inquiry into the Nature and Causes of the Wealth of Nations. London: W. Strahan.
Spence, M. (1973). Job market signaling. The Quarterly Journal of Economics, 87, 355-374.
Stewart, N., Chater, N., & Brown, G. D. (2006). Decision by sampling. Cognitive Psychology, 53, 1-26.
Sugden, R. (2003). The logic of team reasoning. Philosophical Explorations, 6, 165-181.
Tenenbaum, J. B., Kemp, C., Griffiths, T. L., & Goodman, N. D. (2011). How to grow a mind: Statistics, structure, and abstraction. Science, 331, 1279-1285.
Tomasello, M. (2008). Origins of human communication. Cambridge: MIT press.
Tsetsos, K., Usher, M., & Chater, N. (2010). Preference reversal in multiattribute choice. Psychological Review, 117, 1275.
Tversky, A. (1972). Elimination by aspects: A theory of choice. Psychological Review, 79, 281-299.
Vlaev, I., Seymour, B., Dolan, R. J., & Chater, N. (2009). The price of pain and the value of suffering. Psychological science, 20, 309-317.
Wellman, H. (1990). The Child's Theory of Mind. Cambridge MA: MIT Press.
Wolpert, D. M., & Landy, M. S. (2012). Motor control is decision-making. Current Opinion in Neurobiology, 22, 996-1003.
The large literature on the instability and frailty of human judgment and decision making (e.g., Kahneman & Tversky, 2000) mentioned above creates challenges for the entire rational choice approach, in both cognitive science and economics. Here, though, we focus on the complementary question of how rational choice approaches may be shared between economics and cognitive science.
