Why Natural Selection cannot Explain Rationality Elias L. Khalil1 ABSTRACT Biologists recognize that organisms adjust choice when constraints change, so-called “phenotypic plasticity.” Economists call it “rationality.” But what is the origin of rationality? Neo-Darwinists conceive rationality as a trait. But this cannot be the case. Let us suppose two lineages of rationality, R1 and R2. Natural selection would supposedly favor R1 over R2 under C1 constraints and vice versa under C2 constraints. However, if agents are using different rationalities, the fitness functions are incommensurable. For them to be commensurable, there must be only a single kind of rationality, R. But how could R=R1 and R=R2, when R1R2? Keywords: selection optimization; rationality optimization; phenotypic plasticity; rationality- qua-trait; rationality-qua-method; Organismus economicus; Organismus automaton 1 This article shares the subtitle with a working paper, Max Planck Institute of Economics, #2006- 22. This article was supported by the Konrad Lorenz Institute for Evolution and Cognition Research. Older versions received comments from Richard Posner, Ulrich Witt, Gerhard Müller, Werner Callebaut, Steven Orzack, Steve Abedon, Jack Vromen, Brian Charlesworth, Gordon Tullock, Timothy Crippen, Michael Ghiselin, Howard Margolis, Robert Axelrod, Richard Levins, Richard Nelson, Joseph Lopreato, R. Preston McAfee, J.S. Metcalfe, Peter Taylor, Elliott Sober, Stanley Salthe, Casey Mulligan, Franz Weissing, Brian Skyrms, Yew-Kwang Ng, Paul Griffiths, Avi Waksberg, Martin Burd, Ellen Larsen, Deby Cassill, and Michael Dunstan. The usual caveat applies. 0 1 1. Introduction What does an amoeba do on an average day? It simply cannot afford to sit around and hope that some manna will fall from heaven. Nor can it, given the high cost of motion, afford to roam the neighborhood randomly in the hope of encountering some yeast. The amoeba must make efficient decisions: When it moves in search of nutrients by using its “false feet,” it moves via chemotaxis carefully towards higher food concentration gradients and away from (with some slight exceptions) toxic environments. However, when starved, these unicellular, solitary organisms have a strong incentive, or what biologists and psychologists call “signal” or “stimulus,” to undertake an appropriate response: Economists call such a response, when performed by a number of organisms together, “collective action.” When about 104–105 amoeba act together in such a fashion, they form something equivalent to what economists call a “club” [Eichinger et al., 2005], as in the sense of a “club” or a “society” (but not an organism) [Strassmann et al., 2000; Schaap, 2007]. The cells that constitute the club, which looks like a slug, undergo differentiation of functions that allows the slug to act as a “bus.” The “bus” affords a more effective transportation to a new environment that would have a higher food concentration. All organisms have to work, and they better work carefully as they search for nutrients, sexual partners, suitable habitats, and safety. All organisms must deliberate and choose among alternatives. They face an environment that is neither uniform across space nor static across time. All organisms must make decisions, and make the best decisions possible given that each decision involves a tradeoff. For neoclassical economics, the tradeoff entails that every choice, when a resource is fungible, involves an opportunity cost. To migrate to another region might 2 afford more protection, but it involves an opportunity cost that can be unjustified by the meager marginal protection. If the making of the best decision possible allows us to call human agents “Homo economicus,” it should equally allow us to call other living agents “Organismus economicus” or, in the above case, “Amoeba economicus.” The idea that organisms are rational, called the “Organismus economicus” hypothesis, stands in stark contrast to the “Organismus automaton” hypothesis. The latter stipulates that the observed decision making by organisms is, in the final analysis, an illusion. Organisms behave according to programs that are ultimately hard-wired. Organisms do not make choices. They simply act to hard-wired repertoires. At the face of it, no biologist of any kind subscribes to the Organismus automaton hypothesis. Biologists clearly recognize that behavior is the result of interaction between genetic make-up and environmental clues, i.e., incentives. To wit, biologists have diligently shown that phenotypic plasticity is virtual to all taxa and all kingdoms of life [Gould, 1977; Stanley, 1979; Matsuda, 1987; West-Eberhard, 1989; Raff, 1996; Müller & Newman, 2003; Hall et al., 2004]. Organisms develop and behave in light of environmental constraints, as well as in light of inherited constraints known as traits. So, why does this paper invoke the Organismus automaton hypothesis, to which no biologist or, at least, no contemporary biologist subscribes? The hypothesis is used here as the baseline to show an inconsistency. If biologists recognize phenotypic plasticity, and phenotypic plasticity is nothing but rationality, why do many biologists and most social scientists shun away from the Organismus economicus hypothesis, i.e., from upholding the view that organisms make rational decisions? 3 The concept is involved. This paper, given its limited question, need not digress into the wide literature on rationality and optimization in philosophy, evolutionary biology, psychology, and economics [e.g., Nozick, 1969; Maynard Smith, 1978, 1982; Danielson, 1998; Dupuy, 1998; Elster, 2000; Smith, 2003; Khalil, 2007b,c]. This paper rather takes the Organismus economicus hypothesis (rationality) as a non-problematic proposition. This is far from the case, though. In light of behavioral decision research in psychology and behavioral economics, agents exhibit some deviations from rationality in some situations [Camerer et al., 2004; Fudenberg, 2006; Pesendorfer, 2006]. This has been interpreted to support the thesis that agents are “quasi- rational” [e.g., Thaler, 1992, 1994; Jolls et al., in Sunstein, 2000, pp. 13-58]. It also has been interpreted to support the thesis that agents are first and foremost creatures of habit, procedure, and routine [e.g., Gigerenzer, 2005, 2006; see also Goldstein et al., 2001; Rieskamp et al., 2006]. The procedural/routine interpretation regards deviations from rationality as not indicative of irrationality or quasi-rationality. The deviations are rather indicative of the fact that preferences are often vague or underdetermined and, hence, are partially determined by one’s accidental encounters or experiences. This interpretation actually goes back to Herbert Simon’s [1977] theory of “procedural rationality” and even beyond, to American pragmatic philosophy [see Mousavi & Garrison, 2003; Khalil, 2008b]. Such experimental findings, whether used to support the irrationality thesis of Richard Thaler or the procedural rationality thesis of Herbert Simon, have made inroads into the toolkit of a few ethologists [e.g., Marsh & Kacelnik, 2002]. Such deviations are often likened to optical illusions. Note, optical illusions do not mean that the sense of sight is dysfunctional. Likewise, to push the analogy further, deviations from rationality do not mean that agents are irrational. To wit, deviations from rationality actually 4 confirm the principles of rationality [Khalil, 2007b,c]. To be precise, there are actually unrelated kinds of deviations from rationality [Khalil, 207e]. In any case, these details and debates should not influence the question pursued in this essay. Sill, what do we need to know about rationality for the purpose of this article? Briefly, there are two definitions of rationality, the “technical” and the “action” definitions [Khalil, 2007a]. The technical definition is about consistency [Kreps, 1990], which involves two major axioms: Are the agent’s preferences well-ordered (the transitivity axiom) and do they cover all the bundles or choices in the choice set (the completeness axiom)? Such axioms stipulate only the necessary conditions of rationality. Still, why would the agent behave in a way to enhance wellbeing? We obviously need a sufficient condition. Namely, we need the action definition: Rationality is about action in the direction of enhancing wellbeing [e.g., Becker, 1976, ch. 1]. Such action involves deciding among alternatives. The decision is needed given that resources are scarce and the ability of the organism, as defined by its internal constraints (traits), is limited. According to the action definition, the decision of the organism would change in response to the change of external and internal constraints. Note, one such internal constraint is intelligence. Many investigators [e.g., Ng, 1996] confuse rationality with intelligence. Such confusion is avoided here. Intelligence is merely a mental power not different from physical power or other traits, while rationality is about taking decisions that maximize wellbeing given the traits and external constraints. Put differently, rationality is not a constraint. It is about the response to changing constraints such as intelligence. Going back to our question, why do biologists subscribe to phenotypic plasticity—while tend to avoid the postulate that organisms are rational agents? Such an inconsistency, though, is 5 no longer hidden. Under the sustained pressure of the ubiquity of phenotypic plasticity, biologists are under pressure to explain the determinants