Among the mechanisms of epistemic and practical cognition are some we recognize as reasoning. The reasoning involved in epistemic cognition has generally been called " theoretical reasoning", but that label is somewhat misleading, so I will call it epistemic reasoning. The reasoning involved in practical cognition is practical reasoning. I will refer to epistemic and practical reasoning jointly as ratiocination.4 Without yet trying to characterize reasoning, we can nonetheless recognize that there are other cognitive mechanisms that also produce beliefs and direct activity. This is required because reasoning is slow. Many
4 In Pollock [1986 and 1989], I used the term "intellection" instead of "ratiocination".
aspects of reasoning are essentially serial. Whenever possible, human cognition is accelerated by employing our inherently slow hardware in parallel processing. Unfortunately, much of reasoning cannot be done in parallel, so human cognition includes many nonratiocinative processes that also issue in beliefs and actions. For instance, a human being does not have to pull out a pocket calculator and compute trajectories in order to catch a baseball or get out of the way of a bus. We have a built-in mechanism that allows us to estimate such trajectories very quickly.
Such nonratiocinative processes have sometimes been called "quick and dirty", but that is a misnomer because they need not be at all dirty. For example, our built-in procedures for computing trajectories are incredibly accurate. Their shortcoming is not inaccuracy but inflexibility. They achieve their speed by building in assumptions about the environment, and when those assumptions fail, the processes may yield wildly inaccurate answers. For instance, if we see that a baseball is going to bounce off a telephone pole, we had best wait until it ricochets before predicting its trajectory. Our built-in "trajectory module" cannot handle this situation accurately, so we use ratiocination to override it temporarily, until the situation becomes one that can be handled accurately by the trajectory module. I will refer to modules like the trajectory module as Q&Imodules ("quick and inflexible").
It must not be supposed that human Q&I modules are concerned exclusively with motor skills. Psychological evidence strongly suggests that most everyday inductive and probabilistic inference is carried out by Q&I modules.5 In this case, the modules really are rather dirty. Accurate probabilistic reasoning is in many cases computationally infeasible, and so humans appear to rely upon processes like Tversky's "representativeness heuristic", which often yield results incompatible with the probability calculus.6 This is not to say, however, that it is unreasonable to rely upon such approximation methods. The alternative of explicit reasoning is too slow for many practical purposes, and rough approximations often suffice for the purposes for which we need the probabilities.
Much ordinary inductive inference is also the result of Q&I modules. Inductive inference can be carried out ratiocinatively—by explicit reasoning from explicitly stated data—and scientists try to do that. But
5 Tversky and Kahneman . 6 Tversky .
this requires the storage and processing of huge databases, which is exhaustive of system resources and computationally expensive. Only in science do we tend to accumulate a large body of data, peruse it, and then engage in explicit inductive reasoning about it. In ordinary life, we tend to employ procedures that allow us to use the data as we acquire them and then forget the data, forming provisional generalizations that we modify as we go along. This is much more efficient, but it is subject to pitfalls having to do with the possibility of non-independent data that cannot be judged non-independent because they are no longer recalled.7
Thus far I have talked about Q&I modules only for belief formation, but Q&I modules are equally important in practical cognition. In addition to the practical Q&I modules that will be discussed below, I suspect that in at least some cases, emotions constitute Q&I modules for practical reasoning. For instance, being afraid of tigers initiates quick avoidance responses without our having to think about it—a very useful reaction for anyone who is likely to encounter tigers unexpectedly. Embarrassment, indignation, and so forth, may similarly be practical Q&I modules whose purpose is to supplement explicit practical reasoning in social situations. This provides a computational role for these emotions and throws light on why human beings are subject to them. This account is not applicable to all emotions. For instance, it seems unlikely that depression can be viewed as a practical Q&I module. But perhaps what this shows is that what have been called "emotional states" constitute a grab bag of different kinds of states playing different roles in human beings.
The advantage of Q&I modules is speed. The advantage of ratiocination, on the other hand, is extreme flexibility. It seems that ratiocination can in principle deal with any kind of situation, but it is slow. In complicated situations we may have no applicable Q&I modules, in which case we have no choice but to undertake explicit reasoning about the situation. In other cases, human beings accept the output of the Q&I modules unless they have some explicit reason for not doing so. Ratiocination is used to monitor the output and override it when necessary. In sum, the role of ratiocination should be (1) to deal with cases to which Q&I modules do not apply and (2) to monitor and override the output of Q&I modules as necessary. A rational agent can be viewed as a bundle of Q&I modules with ratiocination sitting on top and tweaking the output as necessary.
7 This will be discussed further in chapter 2.
A cognitive agent could be devoid of ratiocination. Its cognition could consist entirely of Q&I modules. It is quite possible that many moderately sophisticated animals work in this way. I will take the defining characteristic of rationality to be the supremacy of ratiocination over Q&I modules. A fUlly rational agent is one in which ratiocination has total overall control. This is not to denigrate the importance of Q&I modules. It is doubtful that any agent could survive in a hostile world relying upon ratiocination as its exclusive means of cognition, but what sets rational agents apart from others is that ratiocination provides the court of last appeal. Given time to make the appeal, the agent will direct its activity in accordance with the pronouncements of ratiocination insofar as these differ from the pronouncements of its Q&I modules. We can divide rationality into epistemic rationality and practical rationality. In a fully epistemically rational agent, epistemic reasoning takes ultimate precedence over Q&I modules in belief formation, and in a fully practically rational agent, practical reasoning takes ultimate precedence over Q&I modules in directing activity. It will become apparent that human beings are closer to being fully epistemically rational than they are to being fully practically rational.
Was this article helpful?
Since World War II, there has been a tremendous change in the makeup and direction of kid baseball, as it is called. Adults, showing an unprecedented interest in the activity, have initiated and developed programs in thousands of towns across the United States programs that providebr wholesome recreation for millions of youngsters and are often a source of pride and joy to the community in which they exist.