A cognition paradigm clash: Simon, situated cognition and the interpretation of bounded rationality

Abstract

Simon’s notion of bounded rationality is deeply intertwined with his activity as a cognitive psychologist and founder of so-called cognitivism, a mainstream approach in cognitive psychology until the 1980s. Cognitivism, understood as ‘symbolic information processing,’ provided the first cognitive psychology foundation to bounded rationality. Has bounded rationality since then fully followed the development of cognitive psychology beyond symbolic information processing in the post-Simonian era? To answer this question, this paper focuses on Simon’s opposition during the 1990s to a new (paradigmatic) view of cognition called situated cognition, which has since put into question the entire view in cognitive psychology of humans as symbolic information processors. This paper then reads the cognitivism/situated cognition debate through the lens of current bounded rationality research in economics, in order (i) to inquire into whether it has tackled the issues in that controversy; (ii) to envisage possible new foundations for a cognitive psychology-based bounded rationality.

Keywords:

• Herbert A. Simon
• bounded rationality
• situated cognition theory
• economics and cognitive psychology

Acknowledgments

A special thanks goes to Antonio Mastrogiorgio for the discussions on the importance of cognitive psychology for economics. Previous versions of this paper greatly benefited from remarks by Stefano Fiori, Nicola Giocoli, Ivan Moscati, Roberto Scazzieri, Carlo Zappia, and by the anonymous referees. I also thank the participants and discussants at the conferences ‘Economics and Psychology in Historical Perspective’ (Paris), ESHET 2015, STOREP 2015, AISPE 2016. To Profs. John B. Davis and D. Wade Hands goes my gratitude for their courtesy. The usual caveat applies.

Notes

1. According to the American Psychological Association (APA), cognitive psychology concerns ‘the study of higher mental processes such as attention, language use, memory, perception, problem solving, and thinking’ (see the Glossary of Psychological Terms at http://www.apa.org/research/action/glossary.aspx).

2. Given the extremely extended literature on this new turn associated with terminological subtleties (still in need of stabilization), a rich encyclopedia entry could be a good starter (Wilson & Foglia, 2011). For the use of the notion of ‘paradigm’ in this context see, e.g. Shapiro (2011), and critically Wheeler (2014). Roth and Jornet (2013) provide a useful concise history. Useful handbooks are Calvo and Gomila (2008) and Shapiro (2014). See also Wallace, Ross, Davies, and Anderson (2007) for an even wider view of the ‘post-cognitivist’ panorama.

3. Even before entering into definitions of these terms, it is important to note that the notion of ‘distributed cognition’ as intended here is not the same thing as the connectionist approach to cognition that has had a limited but recognizable place in bounded rationality in economics (see section 5 below).

4. For some historical context regarding Simon’s activity as a cognitive psychologist, see Sent (2001), Crowther-Heyck (2005) and, of course, Simon (1991). Traditionally, Simon (1955) is the work for which he is mostly known to economists, but it historically belongs to a previous period with respect to Simon’s full development of ‘cognitivist’ cognitive psychology.

5. For a more general reference on the ‘road-not-taken’ methodology, see, e.g. Leijonhufvud (2006).

6. Gigerenzer, Todd, and the ABC Group (1999, p. 13) also acknowledge that Simon was sometimes inconsistent in his definition of bounded rationality.

7. The consistency of Kahneman and Tversky’s thought with Simon’s framework has sometimes been put into question, simply because Simon did not see himself as a true Simon₂ theorist. Nonetheless, as Kahneman and Tversky themselves clearly remark, their results ‘are consistent with the conception of bounded rationality originally presented by Herbert Simon’ (Tversky & Kahneman, 1986, pp. S272-S273).

8. Fiori (2011) also takes into consideration the reconceptualizations of the notion of bounded rationality as developed by James March and by Nelson and Winter. Davis (2010) identifies another strand of bounded rationality research in Lichtenstein and Slovic’s (1971) program of endogenizing preferences. In the course of the discussion, selective reference to these other strands will be provided.

9. Properly understanding the role of modularity is fundamental because it allows us to understand statements like ‘it is of utmost importance to understand that [Simon does not] promote a global unified computational model of the mind’ (Mirowski, 2002, p. 533). Modularity, in fact, casts a global view on cognition but, at the very same time, allows the operational construction of specific models of human cognition focusing on selected cognitive functions.

10. The interference between human cognition and artificial intelligence, mediated first through the computer metaphor and then through actual computer programs, is what makes cognitivism an instance of ‘cyborg’ science. In this picture, Simon is (maybe sharing this place with von Neumann) the most important cyborg scientist contributing to economics (see Mirowski, 2002, ch. 7).

11. ‘Herbert’ was the name of one of Brooks’ robots (note the ironic reference to Simon’s first name. Another robot was ironically labeled ‘Allen,’ after the name of Simon’s main collaborator, Allen Newell; see Brooks, 1990). Herbert’s task was to collect soda cans in a laboratory environment, and it accomplished this task in an apparently intelligent way. Nonetheless, its functioning principle was not representationist or symbolic: Herbert did not rely on any map of the laboratory and neither did it chart courses for picking up the cans. It was simply provided with a sensor to detect can-like shapes, along with a locomotion system relying on impact-preventing sensors.

12. Other debates would focus on more specific topics, such as, for instance, learning transfer (Anderson, Reder, & Simon, 1996, 1997; Greeno, 1997). These debates testify that Simon never reconciled with SC, the distance between them remaining sidereal.

13. Cognitivism had always put a decisive emphasis on the objective of unifying the other theories of cognition. It never conceived itself as just one theory of cognition among others. This was, in a sense, its distinctive trademark (see, e.g. Newell, 1990).

14. The link between bounded rationality and cognitivism was repeatedly asserted by Simon. One becomes aware of this even by cursorily skimming through his collected papers (Simon, 1979–1989, 1981, 1982; Simon, Egidi, Marris, & Viale, 1992).

15. Claiming that even Gibson’s affordances could be considered part of a symbolic approach (Simon, 1990) is symptomatic of Simon’s pan-cognitivist attitude.

16. See, e.g. Kao and Velupillai (2015), which tracks the research thread that follows Simon’s traditional computational approach to bounded rationality.

17. ‘Codification’ is a different notion with respect to ‘processing.’ Paivio (1986) introduced the theory of dual-coding, according to which representations could be coded either in a linguistic mode or in an image-based mode. In either case, we talk about symbols in Simon’s sense.

18. As Vera and Simon bluntly put it: ‘the fact that many mental processes are undoubtedly unconscious or subconscious says nothing about whether these processes are symbolic or not’ (Vera & Simon, 1993, p. 10).

19. In the heuristics and biases approach, symbols undergo an associative process that is typically not hierarchic as in classical cognitivism (Morewedge & Kahneman, 2010).

20. Another feature of System 1 is ‘tacitness’ (see Evans, 2008). The tacit character of bounded rationality has been emphasized by March (1978). The same ‘orthogonality argument’ valid between consciousness and symbolic codification applies to the relationship between tacitness and symbolic codification.

21. The notion of adaptation in the ecological rationality approach appears to be less interactionist than Simon’s that was, on its turn, more restricted than cyberneticists’ notion (see Crowther-Heyck, 2005, p. 193).

22. Vernon L. Smith, however, refers, at least rhetorically, to Simon’s cognitive psychology, when he claims: ‘[w]hat is the subject’s perception of the problem that he or she is trying to solve? This is the legacy of Simon that is prominent in the methodology of experimental economics […]’ (2007, p. 40).

23. A deeply ‘situated’ research program in decision-making is the ‘naturalistic decision-making’ research program, i.e. the study of real-world decisions (Klein, 2008). Damasio’s somatic marker hypothesis (Damasio, 1994) has recently provided a programmatic link between the brain and body in neuroeconomics through the notion of emotions (Loewenstein, Weber, Hsee, & Welch, 2001: see Simon, 1967 for the cognitivist approach to emotions). Some research that specifically links ‘bounded rationality’ with the SDE literature is Spellman and Schnall (2009) and Mastrogiorgio and Petracca (2016).

24. In this regard, even the Simonian cognitivist topic par excellence, i.e. simulation, is today granted centrality in the embodied framework. However, here simulation is no longer understood as simulacrum of human cognition. Foundational human faculties, such as ‘understanding,’ would work through embodied simulation, i.e. through the re-enaction of sensory-motor experience (e.g. each time one faces the notion of ‘cold’ one simulates, i.e. partially activates, the area of the sensory-motor system related to temperature perception: this is a fundamental requirement for understanding the notion of ‘cold’ in all its instances) (see Gallese & Lakoff, 2005).