This photo shows Betty the crow retrieving a bucket containing meat from a well, with a wire she has just bent. Image courtesy of Behavioural Ecology Research Group.
My concern in this section is with the philosophical usage of the term "rationality", as opposed to the broader usage by economists (who emphasise consistency of choice in maximising "utility", regardless of the process and the goal) and biologists, for whom rationality is the consistent maximisation of inclusive fitness across a set of relevant circumstances (Kacelnik, 2004). The standard philosophical definition of rationality emphasises the process by which decisions are made: rational beliefs are arrived at by reasoning, and "rational beliefs are contrasted with beliefs arrived at by emotion, faith, authority or arbitrary choice" (definition of "Rationality" by H. I. Brown in Oxford Companion to Philosophy, 1995, p. 744, cited in Kacelnik, 2004). Although the economic and biological behaviours described by Kacelnik (2004) could be called "rational" in the sense that they describe outcomes that a hypothetical rational agent would choose, they can also be generated by natural processes that can be fully described using non-rational terminology. Even the alleged ability of animals (Hurley and Nudds, 2004) to follow certain logical rules (e.g. exclusion rules, transitivity) which characterise elementary reasoning, may not require a mentalistic explanation: other, less "cognitive" explanations in terms of associative conditioning have been proposed (Allen, 2004b). In keeping with the methodology proposed for investigating mental states, I shall therefore refrain from ascribing rationality to biological systems and economic agents unless doing so generates better scientific predictions about their behaviour than those made by alternative, less "cognitive" accounts.
The standard philosophical definition of rationality is not without difficulties. At first glance, it appears circular: rational agents are those that adopts beliefs on the basis of reasons. Additionally, as Kacelnik (2004, p. 10) points out, the processes by which human beings arrive at their beliefs are often inaccessible to introspection, making it impossible in all but the simplest cases to establish whether their beliefs are truly rational. Applying the philosophical definition of rationality to non-human animals is even more problematic.
My aim here is twofold: to assess the merits of what I consider to be the best philosophical argument against the possibility of rationality in non-human animals, and to formulate a definition of rationality which is philosophically robust but applicable to non-human animals.
The argument I discuss here was articulated by Kenny (1975) and is discussed at length in Leahy (1994, pp. 154-156). The underlying assumption is that rationality pertains to means and ends and is distinguished from other forms of end-directed behaviour - such as the intentional agency I described in chapter two - by the agent's grasp of the connection between the means and the end. Kenny (1975) cites a passage in Aquinas to explain why non-human animals cannot possess this kind of understanding:
Perfect knowledge of an end involves not merely the apprehension of the object which is the end, but an awareness of it precisely qua end, and of the relationship to it of the means which are directed to it. Such a knowledge is within the competence only of a rational creature. Imperfect knowledge of the end is mere apprehension of the end without any awareness of its nature as an end or of the relationship of the activity to the end. This type of knowledge is found only in dumb animals (quoted in Kenny, 1975, p. 19).Kenny elaborates:
When an animal does X in order to do Y, he does not do X for a reason, even though he is aiming at a goal in doing so. Why not? Because an animal, lacking a language, cannot give a reason... It is only those beings who have the ability to give reasons who have the ability to act for reasons (1975, p. 20).Kenny's assumption that non-human animals lack the kind of language required for justifying their actions remains valid in the light of what we know (see the Appendix to part B). Accordingly, his argument stands or falls on the claim that no animal can be said to act for a reason unless it can give a reason for its actions.
Strictly speaking, a rational act requires an agent to be able to justify its actions to itself, not to others. However, the ascription of rationality to an agent is not warranted unless this self-justification is transparent to outsiders. Kenny's mistake, I suggest, lies in his regarding language as the only tool that can render the agent's self-justification intelligible to others.
On the account I am proposing, rational agency consists in pursuit of a goal which is governed not by a map but by a transformational model, constructed by the agent, which contains only those specific properties of the means that are suitable for realising the desired end, including some properties which the means does not yet possess. If we observe the agent to be fine-tuning the properties of her chosen means over an extended period of time, in a way that transforms the means into something ideal for realising her end, according to a model that we can recognise, then we are justified in calling her behaviour rational.
The Ramseyan "map" metaphor which I used to characterise belief in chapter two is thus inadequate to define rationality: while a minimal map may enable an intentional agent to steer herself home by the shortest route, it does not require the agent to grasp the relevant properties of the means by which she accomplishes this task - including properties that the means does not currently possess.
I should point out that nothing in my definition requires a rational agent's selection of properties that are relevant to attaining an end to be wholly conscious. Rational intuitions, such as those guiding a chess master's selection of the best move, may occur largely on a subconscious level; and there is good evidence (Lakoff and Johnson, 1999) that thought is mostly unconscious. My definition of rational agency is thus perfectly compatible with Kacelnik's observation (2004, p. 10) that our rational thought processes are often inaccessible to introspection.
Case study: rational agency in crows
Perhaps the most impressive example to date of rational agency in non-human animals is that of a crow named Betty, who repeatedly displayed the ability to take a straight piece of wire, craft it into a hook with her beak, and use it to snag a piece of meat in a tube (Weir, Chappell and Kacelnik, 2002; Kacelnik, 2004). She had seen and used supplied wire hooks before but had not seen the process of bending. The crow even used different methods to fashion the hooks on different occasions. The method used by the crow was different from those previously reported and would be unlikely to work with natural materials. She had little exposure to and no prior training with pliant material, and had never been observed to perform similar actions with either pliant or non-pliant objects. The crow's ingenuity appears to surpass anything observed to date in chimpanzees.
Kacelnik (2004, p. 34) warns against attributing rationality to crows, based on a single set of observations: we currently do not know how domain-general the ability of New Caledonian crows to plan and execute solutions to new problems is. Nevertheless, the behaviour of Betty the crow seems to have instantiated what Aquinas referred to as knowledge of the end qua end. Her goal (meat) could only be attained by transforming the shape of the hook so as to make it suitable for snagging the meat, and the care she took to adjust the shape of the hook strongly suggests that she was acting in accordance with a transformational model.
As the vast majority of instances of tool use observed in non-human animals show no such appreciation of the relationship between a means and an end, we can infer that very few animals possess this kind of rationality. Capuchin monkeys in the wild, for instance, use sticks to kill snakes, hit other monkeys, and dig for food, but laboratory tests show that when given a choice of sticks for removing a peanut from a clear Plexiglas tube, they show no insight as to how to use the sticks (for a detailed discussion see Budiansky, 1998a, pp. 122-128).
Besides tool use, rationality in animals may be manifested in other domains as well: Currie (2004) argues that pretence is one clear indication of rationality, and makes a suggestion about the kind of evidence that would justify its ascription to non-human primates. We have to consider the possibility that different animals may manifest different domains of rational agency. As Kacelnik (2004, p. 34) points out, even for humans there is no such thing as totally domain-independent reasoning abilities.