On the origins and consequences of Simon’s modular approach to bounded rationality in economics

Abstract

This paper discusses why in the 1950s Herbert Simon introduced bounded rationality as a modular notion—consisting of a “cognitive” and an “environmental” module—and explores the consequences of this choice. Originally, Simon emphasised cognition in economics and the environment in psychology to meet specific disciplinary interests. Continuing adaptively to emphasise cognition in economics has led, then, to significant unintended consequences: (i) the easier assimilation of Simon’s bounded rationality by neoclassical economics, and (ii) the persistent confusion between Simon’s and Kahneman and Tversky’s contribution. Seeing the recognition of his credit endangered, Simon reemphasised the environment when Gigerenzer introduced environment-based ecological rationality.

Keywords:

• Herbert A. Simon
• bounded rationality
• neoclassical economics
• heuristics and biases
• ecological rationality
• modularity

JEL CODES:

• B21
• B31
• D01
• D90

Acknowledgments

I wish to thank Alberto Feduzi and two anonymous referees for useful and constructive remarks on earlier versions of this paper. Comments from participants and discussants at ESHET 2017 (Antwerp), STOREP 2017 (Piacenza), and AISPE 2017 (Rome) are also gratefully acknowledged. I am also grateful to AISPE for awarding me the ‘Best Paper by a Young Scholar Award’ at AISPE 2017. This work would not have been possible without the invaluable help of Julia Corrin at Carnegie Mellon University Archives. Finally, a special thanks goes to Jean Czerlinski Whitmore, Daniel Goldstein, and Laura Martignon for sharing their recollections and to Gerd Gigerenzer also for the permission to publish extracts from his correspondence with Simon. The usual caveats apply.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Notes

1 Mirowski (2002) points out the “recurrent gaps between what Simon does and what Simon Says” (470).

2 Technically, cognition and environment can be considered as two modules at the same level in Simon’s hierarchical framework. As such, “[t]he loose horizontal coupling permits each subassembly to operate dynamically in independence of the detail of the others” (Simon quoted in Sent 2001, 482, f.10).

3 For the intricate panorama of Simon’s contracts in the early 1950s, see Crowther-Heyck (2005, 157).

4 Klaes and Sent (2005) reconstruct thoroughly the conceptual history of bounded rationality, emphasising the unstable terminology of the origins.

5 Chester I. Barnard, “Memorandum of detailed observations on Administrative Behavior by Professor Simon,” 24 June 1945, Box 74, Folder 5916 (Herbert Simon Papers, Carnegie Mellon University Archives, hereafter abbreviated as HSP). See also Crowther-Heyck (2005, 132).

6 The memorandum was presented, after solicitation by Marschak, at Cowles in two seminars on 15 and 16 November 1951. The November 16 “staff meeting,” reserved to the inner circle of economists, had the eloquent title: “The Strategy of Model Construction: Relaxing the Rationality Principle” (Letter from Jacob Marschak, 25 October 1951, Box 41, Folder 3186, HSP).

7 Simon received a negative feedback on this strategy from Ward Edwards, who said “I doubt whether it makes good strategy to restrict the applicability of any model so severely as I feel you do” (Letter from Ward Edwards, undated, 1954, Box, 97, Folder 8456, HSP).

8 Originally published as a RAND paper, RAND P-365, 20 January 1953.

9 The manifesto of Simon’s research at Ford is “Research into Behavior in Organizations” (see Ford Foundation Working Paper #1, co-authored with Harold Guetzkow, 28 February 1952, Box 16, Folder 1138, HSP). The most noticeable output of Simon’s research in this context would be the well-known Organizations (March and Simon 1958).

10 The final version builds upon the draft “Rational Choice and the Structure of the Environment II: The Difficulty of the Environment,” (22 April 1954, Box 63, Folder 4854, HSP). In that paper, Simon generalised the notion of environmental “difficulty”.

11 It is hard to find mention after 1952 of the environment or of the 1956 paper in Simon’s correspondences with leading Cowlesmen, such as Koopmans and Marschak. Interaction with them mostly concerned the progress of the 1955 paper (see, e.g., Letter to Jacob Marschak, 11 April 1953, Box 41, Folder 3186, HSP; Letter to Tjalling Koopmans, 9 March 1953, Box 41, Folder 3185, HSP).

12 The original RAND P-365 did not present appendixes. The discussion on chess was originally sketched within the text and then expanded in a version circulated at the Carnegie Institute of Technology (Simon 1998a, 67). This latter version contained three appendixes the content of which, as Simon wrote to Marschak, “represents largely work of the past six months under the ONR [Office of Naval Research] contract” (Letter to Jacob Marschak, 9 July 1953, Box 41, Folder 3186, HSP).

13 Simon was always well aware of the disciplinary target of his articles. As he explicitly said, “[o]n several occasion in the late 1950s and early 1960s I was asked to write survey articles on theories of decision-making in economics and psychology …[Simon 1959], an invited paper for the ‘American Economic Review’, was addressed to economists, while [Simon 1963] was addressed to an audience of psychologists. Consequently, though the two chapters cover somewhat the same ground, their emphases are different” (Simon 1982, Vol.2, 205, emphasis added).

14 For Simon as a founding father of cognitive psychology and AI, see McCorduck (1972). See also Crowther-Heyck (2005).

15 Gigerenzer et al. (1999, 12) recall that Simon, jokingly but tellingly, thought of suing those who misconstrued bounded rationality in terms of constrained optimisation.

16 The case with Stigler is paradigmatic of Simon’s fight on multiple rhetorical fronts. Simon always took the chance to say that he had anticipated Stigler’s model with his 1955 paper (see, e.g., e-mail to Klaus Fiedler, 19 May 1999, Box 82, Folder 6577, HSP), while, at the same time, he openly said to be against Stigler’s “school of search” (e.g., Simon 1979).

17 Referring to his optimising paper of 1951, Simon claims that “most of my theory of the employment contract can be expressed without either equations or maximization. But in that qualitative form it would not have captured the attention of economists, who, in the ‘new institutional economics’, continue to pour the new wine into the old bottle of neoclassical reasoning. At least we have some new wine.” (Simon 1996a, 167).

18 Concessio was at work, for instance, when Simon conceded that utility maximisation could be used to represent a satisficing choice situation while, at the same time, he denied any operational sense to such a representation: “the notion of seeking a satisfactory return can be translated into utility maximising but not in any operational way” (Simon 1959, 262, emphasis added).

19 For instance, in a correspondence with Marschak, we find sentences like “rules for optimal decision making, if they are to be of any real use, must be ‘optimal’ in relation to the information possessed by the decision-maker and the size of the computational problems he can handle,” which did not imply any rejection—if anything the contrary—of the notion of optimisation (Letter to Jacob Marschak, 11 April 1953, Box 41, Folder 3186, HSP).

20 Sent (2000) quotes James March on this, when he said that Simon’s work between 1958 and 1978, “much of it done in collaboration with Allen Newell, […] is best known in psychology and computer science” (389).

21 Compare this statement to Simon’s harsh letter to Rae Goodell (19 February 1979, Box 73, Folder 5773, HSP) who, on the day after the Nobel Prize, had depicted him in a Psychology Today column as an outsider in economics. In the end, Simon synthesised, “If I was an outsider to the economics profession as a whole, I was an insider to its elite.” (Simon 1996a, 326).

22 We can still find traces of the adaptive pressure exerted by economists in a congratulatory letter by Koopmans after the Nobel Prize: “We are hoping, Herb, that you will see this as a challenge to draw economists more into the circle of people you are writing for… [Your audiences] tend to be somewhat insular” (quoted in Mirowski 2002, 266, f. 45).

23 “[In omniscient rationality] all the predictive power comes from characterizing the shape of the environment in which the behavior takes place. The environment, combined with the assumptions of perfect rationality, fully determines the behavior.” (347). In Simon’s characterisation, “[b]ehavioral theories of rational choice – theories of bounded rationality – do not have this kind of simplicity. But, by way of compensation, their assumptions about human capabilities are far weaker than those of the classical theory [of omniscient rationality]”. Simon was here essentially saying that behavioural theories require a focus on the cognitive module even greater than neoclassical theories.

24 Foss (2003) considers those addressing economists as Simon’s “most ‘rhetorical’ papers” (159), and provides a thorough rhetorical analysis. His judgment is that Simon eventually “failed to persuade” (ibid.). Nonetheless, if persuasion partially failed, adaptation was fully accomplished.

25 The Royal Swedish Academy of Sciences, Press Release, 16 October 1978.

26 Ibid.

27 Letter from Barbara J. Redman, 4 January 1978, Box 53, Folder 3930, HSP.

28 “One can hold that the individual will maximize his or her satisfaction as he or she perceives it at the time of decision, subject to the constraints which he or she perceives” (Letter from Barbara J. Redman, ibid., emphasis in the original). In this quote we can see that, for neoclassical economists, individuals are constrained by their perception of the environment not by the environment itself. This demonstrates, once more, that for neoclassical economists adaptation concerned almost exclusively the cognitive module of rationality.

29 For the slightly different categories of “classical” and “modern” behavioural economics, see Kao and Velupillai (2015).

30 For the theoretical points of continuity between Simon and KT, see Petracca (2017).

31 To illustrate, it is telling that in their commemorative article on Simon, Kahneman and Frederick (2004) cite, of all his works, only the unpublished Simon (1935).

32 The proceedings of the conference first appeared in 1986 as a special issue of The Journal of Business. Then, they were collected in Hogarth and Reder (1987).

33 On other occasions, the situation may have been different. Heukelom (2014) provides a letter in which Simon, after a first supportive phase (154), reproached the Russell Sage Foundation for thinking that bounded rationality needed a new approach (159–160). This might be telling of Simon’s fear of being overcome.

34 For instance, Chapter 4 contains, among others, the distinction between “subjectively” and “objectively” rational decisions. Relevant in this context is also the notion of “role,” responsible for the mismatch between objective and subjective characterisations of a choice situation (see Simon 1963, 741–747).

35 It can be said that frames, understood as different structures of decision tasks, represent where the environment proves to be relevant in KT’s framework. See Jullien (2016) for an extensive discussion of this. However, even if the environment plays an important role in KT, it is not understood as a module of adaptive rationality.

36 In the original version of the Nobel Lecture (Kahneman 2002, 458), Kahneman explicitly acknowledged that Simon was among the discoverers of the framing effect. Kahneman, however, specified that Simon’s discovery had happened in the field of problem-solving and not in that of decision-making. That acknowledgment was eventually deleted from the AER version of the lecture (Kahneman 2003).

37 Simon cited abundantly and systematically the paper Kahneman and Tversky (1973) as evidence that individuals do not follow Bayesian statistics in making predictions (e.g., Simon 1976, 1978a, 1978b, 1979).

38 Memorandum “Economics at CMU: An Opportunity for Undergraduate Economics Instruction at CMU,” 19 April 1992, Box 84, Folder 6714, HSP.

39 Letter to George R. Feiwel, 13 February 1984, Box 60, Folder 4517, HSP.

40 Letter to José Luís Cardoso, 6 September 2000, Box 87, Folder 7056, HSP.

41 E-mail to Daniel Kahneman, 8 November 1998, Box 104, Folder 9546, HSP.

42 E-mail from Daniel Kahneman, 9 November 1998, Box 104, Folder 9546, HSP.

43 E-mail to Ariel Rubinstein, 7 February 1997, Box 26, Folder 1849, HSP.

44 Simon refers to Sargent (1993) and Aumann (1997). For Sargent’s and Simon’s notions of bounded rationality, see also Sent (1997). On Simon and game theory, see Sent (2004b).

45 Letter to Spiro J. Latsis, 29 December 1996, Box 86, Folder 6953, HSP.

46 Somewhere else Simon was even more explicit. Commenting on “the recent embrace that bounded rationality has received from people like Tom Sargent and Aumann,” he was eager to add “neither of whom has a clue about what is really involved” (Letter to John Conlisk, 25 July 1996, Box 26, Folder 1835, HSP).

47 The notion of an “armchair method” had long been common among psychology-trained decision scientists (e.g., Edwards 1954, 381).

48 Regarding Arrow, see Arrow (1987), and regarding Selten, see Selten (1990).

49 Letter to Antonio M. Silveira, 1 July 1997, Box 59, Folder 4499, HSP.

50 “Maintaining GSIA’s historical tension: Between the sciences and the professions,” 21 May 1999, Box 62, Folder 4663, HSP.

51 E-mail to Gerd Gigerenzer, 2 September 1999, Box 100, Folder 8799, HSP.

52 Mirowski (2002, 472–473) presents Conlisk very differently as an economist who neglected Simon’s true message.

53 Letter to John Conlisk, 25 July 1996, Box 26, Folder 1835, HSP.

54 “So just as the behavior of an electron will be different, depending on the electrical and magnetic fields in which it finds itself, so the behavior of economic actors will be different depending on the social and historical context in which they find themselves” (Letter to John Conlisk, 25 July 1996, Box 26, Folder 1835, HSP).

55 Daniel Goldstein had been involved in earlier stages of the paper. Personal communication with the authors.

56 Manuscript # 99-035. Simon recommended publication conditional to substantial revisions. Although the article has never been published, parts of it were reabsorbed in Gigerenzer et al. (1999) (personal communication with the authors). As we could not access the submitted manuscript, the content of the paper is reconstructed on the basis of Simon’s referee report.

57 E-mail to Klaus Fiedler, 19 May 1999, Box 82, Folder 6577, HSP.

58 E-mail from Gerd Gigerenzer, 2 September 1999, Box 100, Folder 8799, HSP.

59 E-mail to Gerd Gigerenzer, 2 September 1999, Box 100, Folder 8799, HSP.

60 E-mail from Gerd Gigerenzer, 29 April 1999, Box 100, Folder 8799, HSP.

61 E-mail to Gerd Gigerenzer, 4 May 1999, Box 100, Folder 8799, HSP.

62 E-mail to Gerd Gigerenzer, 2 September 1999, Box 100, Folder 8799, HSP, emphasis added, later published in Gigerenzer (2004, 406). The version that Simon gave to economists is: “At an abstract level, bounded rationality is a very weak theory (almost as weak as SEU maximization with an unspecified utility function)” (Letter to John Conlisk, 25 July 1996, Box 26, Folder 1835, HSP).