Ambrose Bierce, making a satirical joke, defined “Logic” as “The art of thinking and reasoning in strict accordance with the limitations and incapacities of the human misunderstanding” (Bierce Reference Bierce1971, p. 211). The humour depends partly on how the definition confuses the normative aspect of logic, concerned with principles of good reasoning, with a descriptive fact about how people sometimes do reason.
The target article defends a related claim, although without humorous intent: that the study of human reasoning is better pursued as a descriptive enterprise, focused on how people actually think, without reference to normative considerations. However, Elqayam & Evans' (E&E's) own normative proposal (in this regard) demands too much, and its critical target is insufficiently precise.
Imagine someone who reads the entire target article. Before she does so, she says her aim is to figure out what it says. Afterwards, she announces that she has read a detailed argument for the view that normative considerations are an essential part of the study of human reasoning. I suspect that even the authors of the target article would be inclined, with the rest of us, to say that this person had made a mistake, that somehow she had got things wrong. Reasoning does aim at truth, and asking whether it succeeds or fails, or whether a reasoner commits errors, or does as well as he or she could, is always appropriate, if not always easy to answer. The view being defended by the target article requires that these evaluations, and existing discoveries guided by them, be abandoned when we are engaged in scientific study of human reasoning.
Consider, for example, the work of Kuhn (Reference Kuhn1991), based on interviews with people from various professions and walks of life. All were asked to give reasons for their views on unemployment, education, and crime. Analysing the interviews, Kuhn was generously prepared to count as a genuine reason any claim that was “(a) […] distinguishable from description of the causal sequence itself and (b) [bore] on its correctness” (Kuhn Reference Kuhn1991, p. 45). Only 16% of non-college subjects generated genuine evidence for their beliefs about crime, 28% for unemployment, and 29% for education. College subjects did better, but not perfectly (61% for crime, 53% for unemployment, and 66% for education). If we exclude normative considerations, we can't regard Kuhn's work as showing that many people reason poorly. It would, though, be nonsensical to read it descriptively as a discovery about what reasons actually are. Allowing outright irrelevancies and restatements of the conclusion to count as reasons empties “reasoning” of significant content.
E&E appear to recognize this problem when they say, regarding the work of Stanovich and West, that a normativist tendency is “to some extent moderated by the restricted applicability to elite reasoners” (sect. 5.2, para. 3). It is difficult, though, to see how a truly committed descriptivist could think there were such things as elite reasoners, which involves admitting that they are very good at something. Shortly after the “moderated” remark, the target article admits this, rejecting the notion of a reasoning error, since errors imply norms. The descriptive approach demanded here is blind in the same way as a physical study of two clocks, one fast and one accurate, concluding that the each clock consistently follows the laws of physics.
This objection may seem too strong. Don't E&E allow functional considerations, and evaluation of reasoning processes by instrumental lights, but distinguish this from the error of normativism? They try, but don't succeed. They say it is “only when formal systems are regarded as having a priori, unconditional value that the ‘ought’ becomes an evaluative one” (sect. 2, para. 5) and hence an instance of normativism. This sets the bar too high – many (perhaps all) norms of reasoning are not a priori. Most, maybe all, can be understood as conditional, to be used if there is enough time, unless some other criterion trumps them, as giving a fallible justification, and the like.
Other parts of the specification of “normativism” are similarly frustrating. “Normativism” judges reasoning by the standards of a “system,” but “system” isn't defined. This judgement involves “measuring against” and “conforming to,” leaving it unclear whether normativism judges process, or outcome, or both. Normative systems are “formal,” but “formal” isn't defined. In any event, many plausible norms (“repeating the claim isn't giving a reason”) are not formal. In conflict with the claim noted above that it is “only” a priori norms that are problematic (indicating a requirement), we are elsewhere told that “[p]rescriptive normativism can vary according to factors such as the a priori status of the normative system” (sect. 1, para. 8) (suggesting an option). It would help if it was clearer what normativism actually was.
Two points should not need emphasising. First, we do not know whether there is a single ideal theory of reasoning. It seems likely (because individual theories exhibit their own paradoxes and complications) that no one theory could be ideal, and that the best possible reasoning involves a varied toolkit. It does not follow that abstract theories of good reasoning have no role to play in assessing or interpreting actual reasoning. Second, we do not know exactly how our reasoning systems work. It seems beyond doubt that we don't have a single reasoning process, that we can learn a variety of skills and techniques, and that sometimes one of our reasoning dispositions looks by the lights of one theory to be irrational, yet according to another turns out to be an adaptive shortcut or good move. Our computational limits are finite, but except in a few very vague ways, we don't know how to say what kinds of reasoning are definitely beyond us.
The upshot of these two points is that it is genuinely difficult to say in some cases what good reasoning amounts to, and to say either how good actual human reasoning is, or how good it could be. Since “ought” implies “can,” we shouldn't complain of people that they don't do what they can't do. The relevance of any particular norm to human reasoning, then, really is constrained by the limits of our abilities. It does not follow that we should stop caring about the difference between good and bad reasoning.
You have
Access
Ambrose Bierce, making a satirical joke, defined “Logic” as “The art of thinking and reasoning in strict accordance with the limitations and incapacities of the human misunderstanding” (Bierce Reference Bierce1971, p. 211). The humour depends partly on how the definition confuses the normative aspect of logic, concerned with principles of good reasoning, with a descriptive fact about how people sometimes do reason.
The target article defends a related claim, although without humorous intent: that the study of human reasoning is better pursued as a descriptive enterprise, focused on how people actually think, without reference to normative considerations. However, Elqayam & Evans' (E&E's) own normative proposal (in this regard) demands too much, and its critical target is insufficiently precise.
Imagine someone who reads the entire target article. Before she does so, she says her aim is to figure out what it says. Afterwards, she announces that she has read a detailed argument for the view that normative considerations are an essential part of the study of human reasoning. I suspect that even the authors of the target article would be inclined, with the rest of us, to say that this person had made a mistake, that somehow she had got things wrong. Reasoning does aim at truth, and asking whether it succeeds or fails, or whether a reasoner commits errors, or does as well as he or she could, is always appropriate, if not always easy to answer. The view being defended by the target article requires that these evaluations, and existing discoveries guided by them, be abandoned when we are engaged in scientific study of human reasoning.
Consider, for example, the work of Kuhn (Reference Kuhn1991), based on interviews with people from various professions and walks of life. All were asked to give reasons for their views on unemployment, education, and crime. Analysing the interviews, Kuhn was generously prepared to count as a genuine reason any claim that was “(a) […] distinguishable from description of the causal sequence itself and (b) [bore] on its correctness” (Kuhn Reference Kuhn1991, p. 45). Only 16% of non-college subjects generated genuine evidence for their beliefs about crime, 28% for unemployment, and 29% for education. College subjects did better, but not perfectly (61% for crime, 53% for unemployment, and 66% for education). If we exclude normative considerations, we can't regard Kuhn's work as showing that many people reason poorly. It would, though, be nonsensical to read it descriptively as a discovery about what reasons actually are. Allowing outright irrelevancies and restatements of the conclusion to count as reasons empties “reasoning” of significant content.
E&E appear to recognize this problem when they say, regarding the work of Stanovich and West, that a normativist tendency is “to some extent moderated by the restricted applicability to elite reasoners” (sect. 5.2, para. 3). It is difficult, though, to see how a truly committed descriptivist could think there were such things as elite reasoners, which involves admitting that they are very good at something. Shortly after the “moderated” remark, the target article admits this, rejecting the notion of a reasoning error, since errors imply norms. The descriptive approach demanded here is blind in the same way as a physical study of two clocks, one fast and one accurate, concluding that the each clock consistently follows the laws of physics.
This objection may seem too strong. Don't E&E allow functional considerations, and evaluation of reasoning processes by instrumental lights, but distinguish this from the error of normativism? They try, but don't succeed. They say it is “only when formal systems are regarded as having a priori, unconditional value that the ‘ought’ becomes an evaluative one” (sect. 2, para. 5) and hence an instance of normativism. This sets the bar too high – many (perhaps all) norms of reasoning are not a priori. Most, maybe all, can be understood as conditional, to be used if there is enough time, unless some other criterion trumps them, as giving a fallible justification, and the like.
Other parts of the specification of “normativism” are similarly frustrating. “Normativism” judges reasoning by the standards of a “system,” but “system” isn't defined. This judgement involves “measuring against” and “conforming to,” leaving it unclear whether normativism judges process, or outcome, or both. Normative systems are “formal,” but “formal” isn't defined. In any event, many plausible norms (“repeating the claim isn't giving a reason”) are not formal. In conflict with the claim noted above that it is “only” a priori norms that are problematic (indicating a requirement), we are elsewhere told that “[p]rescriptive normativism can vary according to factors such as the a priori status of the normative system” (sect. 1, para. 8) (suggesting an option). It would help if it was clearer what normativism actually was.
Two points should not need emphasising. First, we do not know whether there is a single ideal theory of reasoning. It seems likely (because individual theories exhibit their own paradoxes and complications) that no one theory could be ideal, and that the best possible reasoning involves a varied toolkit. It does not follow that abstract theories of good reasoning have no role to play in assessing or interpreting actual reasoning. Second, we do not know exactly how our reasoning systems work. It seems beyond doubt that we don't have a single reasoning process, that we can learn a variety of skills and techniques, and that sometimes one of our reasoning dispositions looks by the lights of one theory to be irrational, yet according to another turns out to be an adaptive shortcut or good move. Our computational limits are finite, but except in a few very vague ways, we don't know how to say what kinds of reasoning are definitely beyond us.
The upshot of these two points is that it is genuinely difficult to say in some cases what good reasoning amounts to, and to say either how good actual human reasoning is, or how good it could be. Since “ought” implies “can,” we shouldn't complain of people that they don't do what they can't do. The relevance of any particular norm to human reasoning, then, really is constrained by the limits of our abilities. It does not follow that we should stop caring about the difference between good and bad reasoning.