Book Review: Robert Audi, The Architecture of Reason: The Structure and Substance of Rationality. Oxford: Oxford University Press, 2001.
Audi develops analogies between theoretical and practical reason. He takes them to have a similar foundational structure based in experience. He indicates how recognition of this similarity allows insights about one sort of reason to be applied to the other sort. In particular, it becomes clear that egoism is not a reasonable account of practical reason. In addition to these main themes, Audi distinguishes and assesses various sorts of relativism about reason and develops an account of what it is to be a reasonable and rational person. There are numerous interesting points along the way, many more than can be addressed in a short review.
Audi supposes that beliefs may be (defeasibly) grounded in other beliefs or in certain foundational experiences, such as perceptual or sensory experiences, whereas desires may be grounded in other desires and beliefs or in certain foundational experiences of liked or disliked experiences. In both cases the foundations are experiences rather than beliefs about experiences. (He argues that Sellars’ attack on “the myth of the given” does not apply to his version of foundationalism.) He allows for other foundational sources of justification for beliefs in addition to sensory experience, namely, introspection, remembering (both remembering the experience and remembering that something is or was the case), and reason (or reflection or intuition).
Justified non-inferential beliefs are based on foundational sources. Justified inferential beliefs are based on justified beliefs. The same foundational structure holds for desires or wants, where intrinsic wants are analogous to non-inferential beliefs. Foundational experiences for wants are intrinsically liked or disliked because they are pleasant or unpleasant, but this does not mean, say, that wanting to swim for pleasure is wanting to swim as a means to pleasure: one envisions the pleasure as in the swimming. So wanting to swim can be an intrinsic want.
Audi distinguishes objectual wanting (wanting the pain to stop), behavioral wanting (wanting to swim), and propositional wanting (wanting there to be no more war). Wanting the pain to stop is objectual wanting (wanting this to stop) rather than propositional (wanting that I am free from the pain), so in an important sense it is not egoistic. He says that a similar point holds for belief. “Many philosophers have conflated the question of what justifies a belief with the problem of how it can be defended . . . The basis of my justification for believing that there is a tree before me is a particular visual experience . . . I am not part of the object of the experience” (102) and “selfreferential beliefs need not be taken as primary in perception” (103). We should reject epistemic egocentrism just as we reject egoism.
There is at least one disanalogy between theoretical and practical reason. Suppose that P and Q are incompatible propositions, that you have conclusive reasons to believe their disjunction, P or Q, and that your total evidence does not favor one alternative over the other. Then from the point of view of theoretical reason, it would seem you must suspend judgment and not believe one or the other. On the other hand, if A and B are incompatible courses of action (for example, representing two ways of getting to a place you need to get to), if you have conclusive reasons to take one of these courses of action, and if you have no more reason to take one rather than the other, then from the point of view of practical reason you really must decide and must not fail to make a decision. Audi notes this difference but describes it in the following way. In the practical case, you are justified in choosing course of action A and you are justified in choosing course of action B, whereas in the theoretical case you are neither justified in believing P nor justified in believing Q; however, he thinks you might be rational in believing P and rational in believing Q in the case described, which seems absolutely wrong to me.
One worry about Audi’s approach lies in the particular way in which it depends on the notion of a “ground” or “basis” of a belief or desire, whether ultimate or not. He says at one point that one’s grounds for a belief are the sorts of things that “a successful justification for it would provide” where a justification is a process of arguing for the belief by supplying premises that support the belief. But, as already mentioned, one’s ultimate grounds can include experiences, where experiences of the relevant sort are not premises and are not the sort of thing that a successful argument provides. He allows that a justificatory argument might include premises about certain experiences, but it is the experiences, not the claims about them, which are the ultimate source of justification. He allows that the current grounds of a belief need not be the original grounds. The current grounds of a belief support it in the way that pillars support a porch (where one pillar might replace another in bearing the weight). Support is a “kind of psychologically unobtrusive evidential sustenance relation.” However, we cannot identify supporting beliefs with those one would offer if asked to justify one’s belief, because “it can be difficult to tell when we are discovering a new ground . . . and when we are articulating one that was already a tacit basis of that belief.”
Sometimes a belief is “simply retained in memory” with no record of the original reasons and no new grounds. It is not completely clear to me what status Audi attributes to such a belief. Suppose the belief is not currently on one’s mind; it is just sitting there in memory. It therefore cannot be grounded in a present memory experience, because there is no such experience. Is such a belief grounded in memory simply by virtue of being retained in memory? Or is it ungrounded? Some theorists would say that the justification of such a belief depends on its earlier history. But Audi appears to accept a kind of internalism of the present moment which rules out such an appeal to history.
He does say that, “An important exception to the view that intentional actions are grounded in intrinsic desires is this. One might have forgotten why one is going into the kitchen, hence have a desire to do so that is merely non-intrinsic: neither intrinsic nor instrumental.” He calls this a “residual desire” (footnote 8, p. 246). So presumably a belief for which one has forgotten one’s reasons without having acquired any others is also ungrounded, a “residual” belief. At an earlier point Audi says, A “merely non-inferential belief will be neither justified nor capable of conferring justification on any belief grounded in it; for similar reasons I suggest that a merely noninstrumental desire will tend to be neither rational (though it need not be irrational) nor capable of conferring rationality on any desire or action grounded on it” (128).
What worries me about this is that a typical reader of this review may have a million separate beliefs—about what words mean, about the names and phone numbers of acquaintances, about historical dates and various other useful and useless facts. In relation to that vast store of beliefs, the reader’s current experiences are quite limited, including current sensory experiences, current memory experiences, experiences of introspection, and experiences of reflection and intuition. A moment’s reflection suffices to indicate that the actual experiences of the reader at this time are insufficient to serve as any sort of foundational justification of all but a few of those million beliefs.
Furthermore, it is quite possible, even likely, that almost all of those million beliefs are “simply retained in memory” with no record of the original reasons and no new grounds in relevant support relations, although Audi denies it.
Once it is clear that a belief can be inferential by virtue of standing to one or more other beliefs in the kind of psychologically unobtrusive evidential sustenance relation just illustrated, it also becomes apparent that a great many of our beliefs are inferential. They are based on one or more other, evidential beliefs of ours, as opposed to being non-inferentially grounded in a current experience or mental state or simply retained in memory (35).
As we have seen, in a typical case, a person will have a foundation of non-inferential beliefs rational on the basis of experience or reason and a vast superstructure of beliefs based on them (205).
On the contrary, this is far from “apparent”. It rests on a very strong and dubious psychological hypothesis about support relations. To be sure, people do indeed provide justifications for their beliefs when asked, but the justifications often have little to do with why they originally held the views or why they hold them now. People tend to rationalize, they fabricate reasons when asked what their reasons are. When the original reasons for belief are undermined or when actually cited reasons for belief are undermined people tend to continue believing as they do. If challenged, instead of giving up a belief, they come up with other new reasons for it (Ross and Anderson, 1982; Haidt, 2001). Audi notes that, “it is philosophically prudent to try to account for rationality without multiplying beliefs, inferences, or thought processes of any kind beyond necessity” (34). It is also philosophically prudent to avoid unnecessarily speculative psychological hypotheses!
The problem I worry about here applies to many other foundational theories and I am not what the best response to it is. Audi ought to be able to adapt that best response to the attractive theory he presents here.
Haidt, J . (2001). The emotional dog and its rational tail: A social intuitionist approach to moral judgment. Psychological Review. 108, 814-834.
Ross, Lee, and Anderson, Craig A., (1982). Shortcomings in the attribution process: on the origins and maintenance of erroneous social assessments,” in Judgement under Uncertainty: Heuristics and Biases. Cambridge: Cambridge University Press.
Reviewed by Gilbert Harman, Princeton University, 2002.