Tuesday, 27 November 2012

The phylogeny and ontogeny of deductive reasoning

(Cross-posted at NewAPPS)

In a recent paper, the eminent psychologist of reasoning P. Johnson-Laird says the following:
[T]he claim that naïve individuals can make deductions is controversial, because some logicians and some psychologists argue to the contrary (e.g., Oaksford & Chater, 2007). These arguments, however, make it much harder to understand how human beings were able to devise logic and mathematics if they were incapable of deductive reasoning beforehand.
This last claim strikes me as very odd, or at the very least as poorly formulated. (To be clear, I side with those, such as Oaksford and Chater, who think that deductive reasoning must be learned to be mastered and competently practiced by reasoners.) It looks like a doubtful inference to the best explanation: humans have in fact devised logic and mathematics, which are crucially based on the deductive method, so they must have been capable of deductive reasoning before that. Something like: birds had to have fully formed wings before they could fly – hum, I don’t think so… Instead, the wing analogy suggests that there must be some precursors to deductive reasoning skills in untrained reasoners, but the phylogeny of the deductive method (and to be clear, I’m speaking of cultural evolution here) would have been a gradual, self-feeding process.

Now, the first point requiring a clarification is: what do we mean by ‘deductive reasoning’? Depending on how broadly or narrowly we construe the concept of deduction, the question of whether untrained reasoners do or do not engage in deductive reasoning will receive different answers, even on the basis of the same data. And yet, it is surprising that few psychologists have in fact addressed the issue of what they mean by ‘deduction’ in the first place. My own conceptualization of deduction rests on two basic (and fairly uncontroversial) components: (1) the willingness to reason from unknown or false premises; (2) the formulation of indefeasible arguments, where the premises necessitate the truth of the conclusion in that they allow for no counterexamples (i.e. situations where the premises are the case and the conclusion is not). (Oaksford and Chater also insist on the indefeasible vs. defeasible divide, which in turn can be formally treated in different ways, such as from a probabilistic, Bayesian perspective (as they do), or from the perspective of non-monotonic logics, such as in the work of Stenning and van Lambalgen.) Taking these two components, the task is now to explain how they could have emerged, both from the point of view of their cultural phylogeny and from the point of view of their ontogeny in a particular individual. The first question requires a historical approach, while the second question is to be answered on the basis of research in psychology and education.

With respect to the cultural phylogeny of deduction, the hypothesis I am working on at the moment (together with the other members of my ‘Roots of Deduction’ research project) is that dialogical practices of argumentation are the actual historical precursors both for the idea of reasoning from unknown premises and for the concept of indefeasible arguments. It is now widely accepted that the debating, dialectical practices of the early Academy form the background for the emergence of ‘logic as we know it’, which finds in Aristotle’s Prior Analytics its first mature formulation. Now, granting a premise ‘for the sake of the argument’ is a familiar move in these contexts (as can be seen, for example, in Aristotle’s Topics). As for the concept of indefeasible arguments, at this point my working hypothesis is that the requirement of necessary truth preservation was initially a strategic desideratum, a powerful way to ‘beat’ your opponent in a debate. It is only at a later stage that it became a constitutive feature of the deductive method as such.

By emphasizing the argumentative, dialogical origins of deductive reasoning, my account bears some similarities with Mercier and Sperber’s recently proposed ‘argumentative theory of reasoning’, but it differs from their account in two important points: my claim is restricted to deductive reasoning, thus not ruling out that at least some forms of human reasoning may not have dialogical, argumentative origins; my story is a story of cultural evolution, whereas Mercier and Sperber are interested in the biological, evolutionary emergence of reasoning as an adaptation. But to pursue the evolutionary analogy, we may say that the fact that the deductive method was later co-opted as a methodology for scientific inquiry (thus aiming at truth and not just at winning the debate) can be seen as a case of exaptation: a shift of function occurred.

With respect to the ontogeny of deductive reasoning, we are now interested in how a human reasoner may develop deductive skills upon training. Again, this training will have to be grounded in cognitive possibilities that are available to humans from the start, but we need not postulate innate deductive abilities to explain how humans can learn to reason deductively. (A helpful analogy here is with writing, as investigated in particular by S. Dehaene: it requires extensive training to be learned, but naturally it taps into skills and abilities which are part of a human’s neural make-up from the start.)

The first component – taking unknown or false premises to reason with – is usually taken for granted by reasoning researchers, but this is arguably a consequence of sampling bias: research on reasoning has for the most part been conducted with university undergraduates, who thus had a fair amount of formal education behind their backs. The few studies with unschooled participants, such as the classical study by Luria, suggest that this too is a skill that needs to be learned, and schooling is the typical context for this to happen (think of how a teacher formulates a simple arithmetic problem by giving the students some initial conditions which they must accept uncritically). But seemingly, there are precursors outside the school context for the practice of taking premises at face value. As investigated by Paul L. Harris (with whom I am collaborating on a research project at the moment) and collaborators, situations of story-telling and pretense play can prompt children to accept premises which they know are false, and thus to reason closer to the deductive canons.

The second component, namely the notion of indefeasible arguments, is where the typical undergraduate participants deviate more clearly from the deductive canons. Here too it seems to be first and foremost a question of training (as shown e.g. by the results in this great paper by Morris and Sloutsky and also in a paper by Evans et al.), especially in school settings. But again, there seems to be a clear precursor in the practice of adversarial argumentation, as suggested by the idea that necessary truth preservation was initially a strategic desideratum in the historical development of the deductive method. (This is a hypothesis to be further investigated, which I intend to do with my collaborators if our funding application is successful. It seems plausible even if one does not hold the view that ontogeny recapitulates phylogeny.)

Now, this is admittedly all somewhat sketchy; there is still a lot of work ahead to spell out these ideas in more detail and provide further corroboration for the main claims. But I hope to have at least convinced some of you that Johnson-Laird’s predicament is only an imaginary one: it is perfectly possible to come up with a coherent story for the emergence of the deductive method in logic and mathematics without having to resort to the prior existence of deductive reasoning in human reasoning practices. (I'm working on a paper on this topic at the moment, so feedback is much appreciated.)

19 comments:

  1. Before attibuting a stupid or naive thought to a writer, it seems to me a sane advice to read what he says in full. Johnson-Laird holds the view that in order to reason , and even to reason well, one does not need logic under the form of deductive rules. His view is that people use a basic technique of mental models. The use of mental models does not involve deductions as applications of formal rules, but the search for counterexamples. So Johnson- Laird suggests that there is a "core" of rationality, which consists in a basic understanding of validity in the semantical sense ( see for instance Johnson Laird and Byrne 1991, 209). He does not make the doubtful inference ascribed to him, as the rest of his paper in fact shows.

    ReplyDelete
    Replies
    1. I'm afraid you are the one who is not quite understanding my main point. Johnson-Laird's mental models account is also a form of 'logicism', just as the mental rules account he criticizes, because both accounts are based on the idea of rationality tied to validity understood as necessary truth preservation (hence the focus on counterexamples). It is this idea, whether semantically or syntactically formulated, that I examine critically in my post.

      Delete
  2. It sounds strange to me to say that the view that there can be reasoning without logic is a form of "logicism" , or that it presupposes a view of reasoning with deductive rules. Or if it is a fom of logicism, then this view is so wide that even someone says that language and meaning are needed to reason is a logicist.
    Johnson Laid has been opposing for years what he calls "mental logic" and the idea that logic goes by deductive rules. But you ascribe to him the idea that logic = deduction. According to him only a general principle of validity is necessary for our pre-theoretical reasoning, but that involves only a modicum of "logic" in the deductive sense. If your claim is that as soon as one has validity in the semantic sense one has deductive rules, that's not JL's view. And he certainly does not claim that one needs to have logic before being able to reason logically.
    Moroever his view is quite compatible with the idea that truth preservation comes from some strategic deliberation . My point was that he precisely did not insist on *deduction* from rules. ( The quote that you borrow from his article actually does not represent his considered view).
    Apart from that , as as interesting can be the revival of the idea that logic comes out of dialectics, I do not see how there can be strategic deliberation without *some* notion of truth. Can " granting a premise for the sake of argument " be accounted for without some notion of the premise being *true* for the sake or argument? Even Aristotle's Topics, as "dialectical" as they can be, use the notion of truth.

    ReplyDelete
    Replies
    1. Two recent publications that use the term 'logicism' in the sense that I am using here: Elqayam and Evans' 2011 BBS article, and Stenning and van Lambalgen's 2008 book. The latter also argue in quite some detail why JL's position is a logicist position despite his opposition to the mental rules approach. Simply put, logic is not only a matter of deductive rules, and in fact in the actual work of logicians, the semantic (model-theoretic) perspective is the more widespread one since the mid 20th century.

      Delete
  3. If one decides that "logicism" has this meaning, fine, I don't want to compete with theoretical Humpty Dumptism. Logicism has had, for more a than a century, another meaning, and it has been associated typically with the reduction of mathematics to logic, as syntactical and proof theoretical conception of logical rules. Certainly semantics and model theory is still a very "logical" view as opposed to the kind of view that has emerged from recent work in cognitive science that you mention. I was just concerned to suggest that :

    1) Johnson Laird is not guilty of the dubious inference that you ascribed to him ( people use logic, therefore they must in some sense have been logical before using logic)

    2) among psychologists he has been one of those , against e.g Piaget and Rips, who have insisted on the fact there there can be valid reasoning without logical rules . So even if he is guitly of keeping a semantic approach,he is one of those who have insisted on the fact that logic does not codify natural reasoning. Is is too committed to the "old" view ? Perhaps. But he was ertainly one of those who did most to criticise it.

    3) one thing is one's conception of logic, another is a theory of how it came about or evolved.

    4) there are degrees in one's opposition to the claim that natural reasoning uses schemes, or patterns or strategies ( I try to be cautious,but may be this would seem to some to be already a commitment to "logicism"?) which resemble what our evolved normative thinking has called "logical rules".

    5) If that opposition goes as far as to tell us that we do not need the notion of truth in our natural reasoning, and that truth does not play any role in the origins of logic. I am very sceptical . But of course I was just expressing my doubts, no more. I still need to be convinced by the emerging consensus that you seem to allude to .

    ReplyDelete
    Replies
    1. One thing that seems to have gotten lost in this debate is that my post was not primarily intended as a criticism of Johnson-Laird's theories specifically. I took his claim to be an invitation to think about how mathematics and logic could have emerged if humans did not engage in deductive reasoning beforehand, and then went on to propose my tentative answer to this question.

      On your point 2), there is nothing that JL is 'guilty' of; he holds a theoretical position with which I happen to disagree, while also acknowledging that he has made decisive contributions, in particular with respect to criticizing the mental rules approach. I in fact also agree with the general idea that everyday reasoning is essentially a 'semantic' affair, for lack of a better term (as I argue in chap. 4 of my recent book 'Formal Languages in Logic'). The main difference between JL's approach and the one I endorse pertains to the divide between monotonic and non-monotonic reasoning, and I am using the term 'logicist' to refer to any position that endorses monotonic reasoning as a normative ideal (also outside the contexts of mathematics and logic).

      Delete

  4. As soon as someones tells me that :

    "It looks like a doubtful inference to the best explanation: humans have in fact devised logic and mathematics, which are crucially based on the deductive method, so they must have been capable of deductive reasoning before that. Something like: birds had to have fully formed wings before they could fly – hum, I don’t think so…"

    I understand that a famous logician and psychologist is confused or viciously circular. I pointed out that he might not be that confused.

    At the end of the post you say :

    "But I hope to have at least convinced some of you that Johnson-Laird’s predicament is only an imaginary one: it is perfectly possible to come up with a coherent story for the emergence of the deductive method in logic and mathematics without having to resort to the prior existence of deductive reasoning in human reasoning practices."

    As I said above, JL agrees that there can be reasoning without logic ( in the deductive sense), so does not need to be convinced. That's what he has ben saying all along.

    Three strands are to be distinguished in what you suggest :

    a) whether everyday reasoning is deductive in the classical sense or non monotonic

    b)whether everyday reasoning can be accounted for with or without the notion of truth

    c) whether everyday reasoning is completely alien to schemes that are typical of logic as it is codified by our textbooks.

    It seems to me that one can answer negatively to each of these questions without being circular or doing bad inference to the best explanation.


    A further point is whether, in studying in particular c) one needs normative ideals. It seems to me that we can't help it. There is a methodological issue here : how can one study logic without a normative ideal ? I think it's not possible to do without . The ideal, though, need not be as strong a what you call "logicism" . But I take it that you disagree.

    But it seems that you agree that some notion of truth has to come into the dialectical picture that you favour, since you say that :

    "My own conceptualization of deduction rests on two basic (and fairly uncontroversial) components: (1) the willingness to reason from unknown or false premises; (2) the formulation of indefeasible arguments, where the premises necessitate the truth of the conclusion in that they allow for no counterexamples (i.e. situations where the premises are the case and the conclusion is not)."

    JL's point was that the search fo couteerxamples is a basic capacity in natural reasoning practices.

    I saw that your point was not to critise JL. My point is not to defend him either, although I would defend the primacy of truth ( against Sperber and Mercier) and the primacy of normative ideals when one discussed issues about how logic evolved. But I agree that this is just handwaving here.







    ReplyDelete
    Replies
    1. I'm feeling uncomfortable with your somewhat aggressive tone hiding behind the anonymity of an alias. I'd much prefer to continue this debate knowing who I am debating with.

      Delete
  5. Replies
    1. :) But I still think that whoever else is reading this discussion (my stats say that this post has had 1200 hits since yesterday) should also know who you are.

      Delete
  6. I wasn't hiding myself at all , and thought that anyone could easily decipher an anagram, but

    ange scalpel = pascal engel

    ( the reason I used this is actually accidental : my gmail account is

    ange.scalpel@gmail.com and the blogger profile asked me to use google blogger)

    Moreover, I am not being agressive , unless you think that asking for more use of the principle of charity is .



    ReplyDelete
    Replies
    1. I don't think there is much point in discussing the rhetoric and ethics of commenting at blog posts, but when on your first comment you used the terms 'stupid' and 'naive' to refer to my characterization of JL's statement (while I used 'odd' and 'doubtful', which to me are of an entirely different nature), my willingness to engage with the (interesting) substantive points you were making decreased a bit.

      But anyway, nothing I say is incompatible with taking truth as a fundamental notion. The question I am addressing is what could have been the narrative leading to the development of the deductive method; it is a genealogical rather than foundational question (perhaps something like the distinction between 'ordre des raisons' and 'ordre des matieres?).

      Delete
  7. Hi Catarina,

    Interesting post. What about the role of analogical reasoning as providing that innate cognitive mechanism which underlies the development of deductive reasoning? It seems to me that the leading models of analogy are sufficiently powerful to do so, although I am not at the moment aware of any literature dedicated to investigating this possibility.

    ReplyDelete
    Replies
    1. I'm not much on top of the literature on analogical reasoning, except for knowing that there's a lot of exciting work being done at the moment. But for me, an important divide is between mono-agent situations and multi-agent, dialogical situations, and my hunch is that analogical reasoning pertains rather to the mono-agent side while deduction (as I argued) pertains to the multi-agent side. Unlike Mercier and Sperber, I do not claim that all reasoning is argumentative (multi-agent) originally; I leave room for aspects of human reasoning not arising from practices of argumentation. So my impression is that much of what is known as analogical reasoning will not have an argumentative origin (even though analogy is also an argumentative scheme).

      Delete
  8. Thanks for the lesson in ethics posting . Some posts would probably be less discussed if they were a bit less hostage to an advertising style, but I suppose it is the usual format of blogs, which is the reason why I very often hesitate to discuss on them, and am clumsy at it. But I feel obliged to have my word when it seems to me that an injustice has been done to an author.

    When one starts a post by saying that a famous psychologist has defended an inference which seems as ridiculous as :

    "Something like: birds had to have fully formed wings before they could fly – hum, I don’t think so"

    my ordinary reaction is to believe that you think that the inference is a bit stupid or naïve, even though you did not use these words. But given that in these topics Dr Pangloss made himself famous, the implicature that Johnson Laird was a bit Panglossian was there. You can cancel it, I agree. The irony of the whole thing is that JL has been for long a stern opponent to Darwinian adaptative thinking about logic.

    Now I am probably mistaken in thinking that you want to prescind from any notion of truth in your account of the origins of logic. But since you mentioned Sperber and Mercier, whose views are - if I understand well - that it's not truth or validity but eristics and dialectical argument which are at the origins of logic, I was wongly supposing that you had similar ideas. Sorry for that. Surely I should have read more of your papers on this.




    ReplyDelete
    Replies
    1. I've also had my fair share of unpleasant exchanges with anonymous commentators, so that's often a reason for me to be a bit hesitant to engage in such debates (I didn't know that you were not particularly aiming at anonymity!).

      But anyway, all is well that ends well, and I appreciate the comments and objections you've raised. As this is a long-term project, there is still lots to do and think about.

      Delete

  9. I shall come back when I shall have done my homework on the articles and books that you suggest . I have read the Sperber - Mercier stuff, and BBS Elqayam and Evans and your comments there. I guess I'm a normativist, this time accepting your label. This promises heated debates or the future,where I shall do my best to behave myself.

    ReplyDelete
  10. I'm coming in late to the debate--maybe not too late?--just for a minor point that touches on the theory Dan Sperber and I put forward.

    First, it says little about the origins of logic as a cultural practice.

    Second, as it stands, it also doesn't say that much about the type of cognitive processes that underpin reasoning -- it is mostly a theory of the computational, ultimate level (although we're increasingly working on the algorithmic level).

    Third, we do not mean to say that truth is foreign to reasoning. When people evaluate other people's arguments, their primary aim is to ascertain whether they are good enough to warrant changing their mind. That reflects an overall concern for truth in some loose sense. It's not all rhetoric (even if it is all dialectic). I'm not sure that Pascal was saying otherwise anyway, but I though it might be useful to clarify.

    ReplyDelete
  11. Dear Hugo

    Thanks for the supervenience ( in the original sense of coming on the top of) !


    Some popular presentations of your view , such as the NYT
    interview from last year emphasize that reasoning is not truth seeking or truth evaluating, but arguing, which goes with
    communication, winning arguments, ascertaining them against opponents and using the techniques codified for centuries under the name of dialectics and rhetoric. Thus ( from the NYT)

    "“Reasoning doesn’t have this function of helping us to get better beliefs and make better decisions,” said Hugo Mercier, who is a co-author of the journal article, with Dan Sperber. “It was a purely social phenomenon. It evolved to help us convince others and to be careful when others try to convince us.” Truth and accuracy were beside the point. "

    Now, there is a strong and a weak reading of this. The strong reading says that truth is outside the picture not only as the purpose or aim of arguing, but also as the dimension of evaluation or assessment of arguments. The weak one says that truth is not the purpose, but that arguing involves evalution of arguments for truth. Clearly when in your BBS paper you emphasize the distinction between production and evaluation of arguments, you grant this.

    I take it that what you say here on the third count is compatible with the weak view . And the weak view seems to me to be compatible with a lot of "classical" accounts of reasoning. Actually it seems hard to argue with the aim of winning an argument without at some point having to evaluate truth . Even the famous Monthy Python story ( " I came here for an argument") registers this : when the arguer receives the knowk of a hammer on his head, he feels that somehing is amiss.


    ReplyDelete