Self-deception, intentions, and contradictory beliefs

Self-deception, intentions, and contradictory beliefs José Luis Bermúdez Philosophical accounts of self-deception can be divided into two broad groups – the intentionalist and the anti-intentionalist. On intentionalist models what happens in the central cases of self-deception is parallel to what happens when one person intentionally deceives another, except that deceiver and deceived are the same person.1 In the classic case of self-deceiving belief formation, the self-deceiver brings it about that they themselves acquire (or retain) a particular belief that p – just as in otherdeception one person intentionally brings it about that another person acquires a belief. In neither case is the (self-)deceiver motivated by conviction of the truth of p. The interpersonal deceiver wants it to be the case that his victim forms the belief that p, while the self-deceiver wants it to be the case that he himself forms the belief that p. Both the self-deceiver and the interpersonal deceiver act intentionally to bring it about that the relevant desire is satisfied.2 According to anti-intentionalist accounts, in contrast, self-deceiving belief formation can be explained simply in terms of motivational bias, without bringing in appeals to intentional action. Self-deception is not structurally similar to intentional interpersonal deception. Whereas intentionalism is generally defended by arguments to the best explanation, anti-intentionalism is often argued for indirectly, through attacking alleged incoherences in intentionalist accounts. Many of the arguments appealed to by anti-intentionalists suffer from failing to specify clearly enough what 1 Some theorists have taken the view that it is pleonastic to talk of intentional deception, since non-intentional deception is a conceptual impossibility (e.g. Haight 1980). This seems wrong. From a purely lexical point of view it seems perfectly possible to describe situations in which there is no intention to deceive as instances of deception. If, for example, one person transmits to a second person a piece of information that they know to be false and that they know will be passed on, then it seems right to say that when the false information is passed on by the second person to a third party an act of deception takes place. One might wonder, however, who is deceiving whom here. Is it the second person deceiving the third? Or is it the first person deceiving the third by means of the second? Only the first interpretation is fully non-intentional, since the second effectively includes the first person’s intention in the description of the episode, even though the first person is not directly involved. There are plausible arguments for both interpretations, but there is no need to arbitrate between them here since intentionalist accounts of self-deception model self-deception on intentional interpersonal deception, rather than on interpersonal deception in general. 2 I discuss further below why this counts as a form of deception. Analysis 60.4, October 2000, pp. 309–19. © José Luis Bermúdez 310 josé luis bermúdez the intentionalist is actually proposing. This paper distinguishes three different descriptions that an intentionalist might give of an episode of self-deception. Separating out the different possible ways in which an intentionalist might characterize an episode of self-deception allows us to evaluate some of the commoner objections that are levelled against the intentionalist approach. I end with a positive argument in favour of intentionalism, suggesting that only intentionalist accounts of self-deception look as if they can deal with the selective nature of self-deception. 1. I am concerned here only with self-deception issuing in the acquisition of beliefs, putting to one side both self-deceiving acquisition of nondoxastic propositional attitudes (such as hope or fear), and cases of retention rather than acquisition. In general terms the intentionalist view is that it is a necessary condition of a subject’s forming a self-deceiving belief that p that they should intentionally bring it about that they believe that p. But there are different ways in which this might happen. For example: (1) S believes that not-p but intends to bring it about that he acquires the false belief that p. (2) S believes that not-p but intends to bring it about that he acquires the belief that p. (3) S intends to bring it about that he acquires the belief that p. Correspondingly, there are three different ways, in ascending order of strength, in which an intentionalist might characterize core episodes of self-deception: (A) Core episodes of self-deception involve intending to bring it about that one acquires a certain belief. (B) Core episodes of self-deception involve holding a belief to be false and yet intending to bring it about that one acquires that belief. (C) Core episodes of self-deception involve intending to bring it about that one acquires a false belief. (A)-type intentionalism attributes intentions of the third type discussed above; (B)-type intentionalism intentions of the second type and (C)-type intentionalism intentions of the first type. It is clear that holding (A) to be true is a bottom-level requirement on any intentionalist account of selfdeception. But (A) does not entail either (B) or (C). (A) does not entail (B), because it is false that one can only intend to bring it about that one believes p when one holds p to be false. One might have no views whatsoever about whether or not it is the case that p and yet intend to bring oneself to believe that p. This would be the position of a complete agnostic persuaded by Pascal’s Wager. Alternatively, one might have evidence for or against p that was too inconclusive to warrant any- self-deception, intentions, and contradictory beliefs 311 thing more than the judgement that p is possibly true, and yet intend to bring it about that one believe that p. Nor does (B) entail (C). I can believe that p is false and intend to bring it about that I acquire the belief that p without it actually being part of the content of my intention that I come to believe a falsehood. That is to say, to put it in a way that parallels a claim sometimes made in epistemology, intentions are not closed under known logical implication. I can know that x entails y and intend to bring about x without ipso facto intending to bring about y. The epistemological parallel is worth pursuing. The thesis that knowledge is not closed under known logical implication can be motivated by placing the following requirements on knowledge (Nozick 1981); I cannot know that p unless (a) were it to be the case that not-p, I wouldn’t believe that p (b) were it to be the case that p, I would believe that p. So, I might know that p entails q and also know that p without knowing that q – because my belief that q, unlike my belief that p, does not satisfy the equivalent of requirements (a) and (b). We can place similar conditions upon intentions. Suppose I know that if I go out for a walk on the moor this afternoon it is inevitable that I will end up killing midges (because they are all over the place and I know that I cannot stop myself from crushing them if they land on me, as they are bound to do). Yet I form the intention to go out for a walk and end up killing midges. Have I killed those midges intentionally? It is plausible that I haven’t. After all, the killing of the midges was not my aim in going out, nor was it a means to my achieving my aim. My aim was simply to go out for a walk, and even if my going out hadn’t resulted in the death of a single midge I would still have gone out. The equivalent of requirement (a) is not met. The case of self-deception seems exactly parallel. I know that my bringing it about that I come to acquire the belief that p will have the consequence that I come to believe a falsehood. Nonetheless, I can intend to bring it about that I believe that p without intending to bring it about that I believe a falsehood – because in a counterfactual situation in which my bringing it about that I believe that p does not result in my believing a falsehood I would nonetheless still bring it about that I believe that p. This reflects the intuitive idea that I do not intentionally bring it about that I believe a falsehood, because that is not my aim. So (B) does not entail (C). Having separated out the three different types of intentionalist position one might still wonder, however, whether they all really count as instances of self-deception. In what sense are subjects who form one or other of intentions (1) to (3) properly described as deceiving themselves? 312 josé luis bermúdez It is pretty clear where the deception comes in (B)- and (C)-type intentionalism, since in both cases the self-deceiver knows that p is false and sets out to manipulate himself into believing that p – just as in intentional interpersonal deception one person sets out to manipulate another into believing something that he (the deceiver) holds to be false. But what about (A)-type self-deception? Things are more delicate here since in (A)-type self-deception the self-deceiver does not necessarily hold the belief they intend to acquire to be false. Nonetheless, they do know that the belief they intend to bring it about that they acquire is not currently warranted by the available evidence. Is this genuine deception? Well, it would count as deception in the interpersonal case. If I intend to bring it about that someone I’m talking to at a dinner-party comes to believe that a particular mutual friend is untrustworthy even though I know that the evidence available to either of us does not warrant that conclusion, then I would properly be described as deceiving them, simply because I would knowingly be contravening the tacit principle that beliefs should be transmitted only when they are held to be true. The case of (A)-type self-deception is exactly parallel. I intend to bring it about that I acquire a belief in ways that are not simply non-standard, but that contravene the norms of truth governing belief formation. I am manipulating my own belief-forming mechanisms. In all three versions of the intentionalist claim, therefore, the deception comes in because the intention is to bring it about that one acquires a belief that one knows one would not have acquired in the absence of that intention. The distinction between self-deception as the intentionalist construes it and wishful thinking (or other comparable modes of motivationally biased belief formation) should be clear. The wishful thinker acquires the belief that p because they want it to be the case that p. The self-deceiver, in contrast, will usually want it to be the case that p – but will also want it to be the case that they believe that p. Moreover, the self-deceiver forms the belief that p as a result of having intended to form the belief that p, whereas there is no such intention in wishful thinking. 2. Let me turn now to what is often taken to be an obvious objection to any intentionalist account of self-deception, but which is actually applicable only to the suggestion in (B) and (C) that the self-deceiver forms an intention to bring it about that he acquire a belief that he thinks is false. There are two main reasons for holding this to be incoherent. First, one might argue that one cannot intend to bring it about that one acquires a belief that one thinks to be false without simultaneously having contradictory beliefs. Yet it is impossible to be in any such state. Second, one might think that the project is guaranteed to be self-defeating. If one knows that the belief is false then how can one get oneself to believe it? self-deception, intentions, and contradictory beliefs 313 The first line of argument is unconvincing. There is certainly something very puzzling about the idea of an agent avowing the contradictory belief that p & not-p. But nothing like this need occur in either (B) or (C), since the two beliefs could be inferentially insulated from each other. It is clear that ‘S believes p at time t’ and ‘S believes q at time t’ do not jointly entail that S at time t has a single belief with the conjunctive content that p & q. So, an account of self-deception can involve the simultaneous ascription of beliefs that p and that not-p without assuming that those two contradictory beliefs are simultaneously active in any way that would require ascribing the contradictory belief that p & not-p. But there is a sense in which this is peripheral, because it is unclear that either (B) or (C) do require the ascription of simultaneous contradictory beliefs. I can start from a state in which I believe that not-p and then intentionally bring it about that I acquire the belief that p without there being a point at which I simultaneously believe that p and believe that not-p. The best way of bringing it about that one moves from a state in which one believes not-p to a state in which one believes p is to undermine one’s reasons for believing not-p and hence to weaken one’s belief in not-p. It seems plausible that one’s confidence in p will be inversely proportional to one’s confidence in not-p. It might be argued that one cannot do something intentionally without doing it knowingly. Hence one cannot intentionally bring it about that one believes p when p is false without knowing that p is false – and so one will have simultaneous contradictory beliefs after all. I shall shortly argue that the premiss is false, but let me concede it for the moment. The first thing to notice is that this only threatens intentionalists who espouse (C). Those who espouse (B) are left untouched. Presumably what one knows is the content of one’s intention and in (B) the content of that intention does not include the falsehood of the belief that one is trying to get oneself to believe. Of course, it is an episode of (B)-type self-deception because one starts off believing that not-p. But one might have forgotten this in the process of tricking oneself into acquiring the belief. But should we conclude that a (C)-type intentionalist is committed to the ascription of simultaneous contradictory beliefs? Not at all. During the process of intentionally bringing it about that one comes to believe a falsehood one will, on the current assumption, know that one is bringing it about that one will acquire a false belief. But during that time one has presumably not yet acquired the false belief in question. So there is no conflict. And when the process of acquiring the belief has come to an end, so too does the intentional activity and the concomitant knowledge that the belief thus acquired is false. Someone might argue that this will still not allow the self-deceived believer to believe that p with impunity – because one cannot believe p without believing that p is true and one 314 josé luis bermúdez cannot believe that p is true if one believes that one has caused oneself to believe it. But this confuses the normative and the descriptive. No doubt one ought not to believe that p if one believes that one caused oneself to believe that p. But as a simple matter of psychological fact people can reconcile those two beliefs. One might believe, for example, that although one initially set out to cause oneself to believe that p the evidence that emerged in favour of p was so overwhelming that one would have come to believe that p regardless. In any case, it seems false that one cannot do something intentionally without doing it knowingly. It is certainly implausible that one might intentionally perform a simple action like pulling a trigger or switching on a light without knowing that that is what one is doing. And it also seems pretty clear that if such a simple action has consequences which one has no means of recognizing or predicting (like assassinating the only woman to believe that Arkansas is in Australia or frightening away a tawny owl) then those consequences are not achieved intentionally. But most intentional actions do not fit neatly into one or other of those categories. Suppose I have a long-term desire to advance my career. This long-term desire influences almost everything I do in my professional life, so that it becomes correct to describe many of my actions as carried out with the intention of advancing my career. Does this mean that I carry them out knowing that they are being done with that intention? Surely not. The intention to bring it about that one acquire a certain belief is closer to the intention to advance one’s career than it is to the intention to switch on a light. As Pascal pointed out, acquiring a belief is a long-term process involving much careful focusing of attention, selective evidence-gathering, acting as if the belief was true, and so forth. It seems likely that the further on one is in the process, and the more successful one has been in the process of internalizing the belief, the more likely one will be to have lost touch with the original motivation. This is not to say, however, that all cases of intentional self-deception will involve unconscious intentions. Far from it. An action can be performed unknowingly even though the intention that gave rise to it was (at the time it was formulated) fully conscious. The point is simply that one can lose touch with an intention while one is in the process of implementing it, particularly when that implementation is a long drawn out process. The fact that an action is precipitated by a conscious intention does not entail that while carrying out the action one remains constantly conscious of the intention that gave rise to it. By the same token, the fact that one is not conscious of the intention while carrying out the action does not undermine the action’s status as intentional. Nor does it threaten the idea that the project is one of deception – any more than in the interpersonal case the fact that I am not at the moment conscious of my self-deception, intentions, and contradictory beliefs 315 original intention to deceive means that my ongoing attempts to manipulate my dinner companion no longer count as deception.3 So, the standard objections to intentionalist accounts of self-deception are less than convincing. This goes some way towards weakening the case for anti-intentionalism, simply because a considerable part of the appeal of anti-intentionalism comes from the puzzles or paradoxes that are supposed to beset intentionalist approaches. But what about the positive case for intentionalism? 3. The positive case for intentionalism is based on inference to the best explanation. The intentionalist proposal is that we cannot understand an important class of psychological phenomena which would normally be labelled self-deception without postulating that the subject is intentionally bringing it about that he come to have a certain belief. It is not enough for the intentionalist to show that such situations sometimes occur – perhaps by citing extreme examples like the neurological patients who deny that there is anything wrong with them, despite being cortically blind (Anton’s syndrome) or partially paralysed (anosognosia for hemiplegia). The intentionalist needs to capture the middle ground by showing that many of the everyday, ‘common-or-garden’ episodes that we would characterize as 3 This shouldn’t be taken to imply that any episode properly described as an episode of deception involves an intention to deceive. As we saw earlier (fn.1), some episodes of interpersonal deception do not directly involve intentions. It is an open question whether an anti-intentionalist account of self-deception can hold episodes of selfdeception to be structurally parallel to those episodes of interpersonal deception that are not intentional. Of course, there is no question of the self-deceiver being divided into three parts. We can think instead in terms of three different temporal stages of a single self-deceiver (with talk of temporal stages not carrying any theoretical loading). But whether there really could be a structural similarity here will depend on which of the two proposed ways of explaining such episodes is adopted. If the first account is favoured, then there seems to be room for a degree of structural similarity between non-intentional self-deception and non-intentional interpersonal deception (although one might wonder whether the concept of deception is not becoming a little stretched). There could be a situation in which the second temporal stage of a selfdeceiver deceives the third stage unintentionally as the result of a course of events set in motion by the first stage. If the second interpretation is adopted, however, then there doesn’t seem to be comparable scope for a structural similarity claim. On the second interpretation, non-intentional interpersonal deception is parasitic on intentional interpersonal deception. So, although the episode itself would strictly speaking only involve a transfer of information between the second and third temporal stages, describing it as an episode of deception would involve bringing in the intention of the first temporal stage. Consequently, if the structural similarity claim were true, purported cases of non-intentional self-deception wouldn’t really count as nonintentional at all. 316 josé luis bermúdez instances of self-deception need to be explained in intentionalist terms. The task is too large to undertake here, but I will make a start on it by trying to show that one sophisticated version of the anti-intentionalist and deflationary approach to self-deception suffers from a fatal defect unless supplemented by an intentionalist explanation. It seems likely that the problem will generalize to any anti-intentionalist strategy. Here are the four conditions that Al Mele (1997, 1998) has deployed to characterize self-deception: (i) The belief that p which S acquires is false. (ii) S treats data seemingly relevant to the truth-value of p in a motivationally biased way. (iii) This motivationally biased treatment non-deviantly causes S to acquire the belief that p, (iv) The evidence that S possesses at the time provides greater warrant for not-p than for p. Examples of the motivational biases that Mele mentions in spelling out the second condition are: selective attention to evidence that we actually possess; selective means of gathering evidence; negative misinterpretation (failing to count as evidence against p data that we would easily recognise as such were we not motivationally biased); positive misinterpretation (counting as evidence for p data that we would easily recognize as evidence against p were we not motivationally biased), and so forth (Mele 1997). This account is radically incomplete, however. What is the connection between our desire that p be the case and our exercise of motivational bias? Where does the motivational bias come from? We can get a sense of how Mele would respond from his (1998) discussion of the model of everyday hypothesis testing developed by Trope and Liberman (1996). The basic idea is that people have different acceptance/rejection thresholds for hypotheses depending upon the expected subjective cost to the individual of false acceptance or false rejection relative to the resources required for acquiring and processing information. The higher the expected subjective cost of false acceptance the higher the threshold for acceptance – similarly for rejection. Hypotheses which have a high acceptance threshold will be more rigorously tested and evaluated than those which have a low acceptance threshold. Mele proposes that, in many cases of self-deception, the expected subjective cost associated with the acquired false belief is low. So, for example, the complacent husband would be much happier falsely believing that his wife is not having an affair than he would be falsely believing that she was having an affair – because he desires that she not be having an affair. So the acceptance threshold for the hypothesis that she is not having an affair will be low, and it is this low acceptance threshold self-deception, intentions, and contradictory beliefs 317 which explains the self-deceiving acquisition of the belief that she is not having an affair. Clearly, then, Mele would say that S’s desire that p be true results in a motivationally biased treatment of data by lowering the acceptance threshold and raising the rejection threshold of the hypothesis that p, thus opening the door to biased treatment of the data. This account is ingenious and provides a good explanation of at least some instances of belief acquisition that one might intuitively classify as instances of self-deception. But there is a fundamental problem that the theory does not seem to address. Self-deception is paradigmatically selective. Any explanation of a given instance of self-deception will need to explain why motivational bias occurred in that particular situation. But the desire that p should be the case is insufficient to motivate cognitive bias in favour of the belief that p. There are all sorts of situations in which, however strongly we desire it to be the case that p, we are not in any way biased in favour of the belief that p. How are we to distinguish these from situations in which our desires result in motivational bias? I will call this the selectivity problem. In response to an earlier presentation of the selectivity problem (Bermúdez 1997), and to a related but different problem identified by William Talbott (1995), Mele (1998) gives the following illustration of how his theory might cope with it. He imagines Gordon, a CIA agent who has been accused of treason. Although Gordon’s parents and his staff of intelligence agents have access to roughly the same information relative to Gordon’s alleged crime, and they all want Gordon to be innocent, they come to different verdicts. Gordon’s parents decide that he is innocent, while his colleagues decide that he is guilty. How can this be, given that the intelligence agents and the parents have the same desire and access to similar information? Mele’s response is that the cost of falsely believing that p is far higher for the intelligence agents than it is for Gordon’s parents. The agents’ desire that Gordon be innocent is trumped by their desire not to be betrayed and their acceptance and rejection thresholds differ accordingly from the threshold of Gordon’s parents (who might be comforted by believing falsely that their son is innocent). The cost-benefit analysis provides a plausible explanation of what might be going on in the Gordon case, but it does not seem to solve the selectivity problem. The selectivity problem is not a problem of how two people in similar situations can acquire different beliefs. It arises, rather, from the fact that possessing a desire that p be true is not sufficient to generate cognitive bias, even if all other things are equal (which they are, perhaps, for Gordon’s parents but not for his subordinates). It is simply not the case that, whenever my motivational set is such as to lower the acceptance threshold of a particular hypothesis, I will end up self-deceivingly accept- 318 josé luis bermúdez ing that hypothesis.4 The selectivity problem reappears. There are many hypotheses for which my motivational set dictates a low acceptance and high rejection threshold and for which the evidence available to me is marginal enough to make self-deception possible. But I self-deceivingly come to believe only a small proportion of them. Why those and not the others? One might suggest that self-deceiving belief formation arises when the self-deceiver possesses, not simply the desire that p be the case but also the desire that he come to believe that p. But the selectivity problem seems to arise once more. Even if one desires both that p be true and that one come to believe that p it is not inevitable that one will form the belief that p. We still need an explanation of why one does in some cases and not in others. Intentionalist accounts of self-deception have a clear and straightforward answer to the selectivity problem. The self-deceiving acquisition of a belief that p requires more than simply a desire that p be the case and a low acceptance/high rejection threshold for the hypothesis that p. It requires an intention on the part of the self-deceiver to bring it about that he acquires the belief that p. The fact that intentionalist theories can solve the selectivity problem in this way seems at least a prima facie reason for thinking that one cannot entirely abandon an intentionalist approach to self-deception. One might add, moreover, that including an intention in the account makes the psychological phenomenon look far more like a genuine case of deception, for the reasons canvassed at the end of §1. Let me end with two qualifications. First, the argument from the selectivity problem can only be tentative, because it remains entirely possible that an anti-intentionalist response to the selectivity problem might be developed. It is hard to see what sort of argument could rule this out. Second, even if sound the argument does not compel recognition of the existence of anything stronger than what I have called (A)-type selfdeception. I have suggested that (B)- and (C)-type self-deception are not conceptually incoherent. But it remains to be seen whether there are situations for which inference to the best explanation demands an analysis in terms of (B)- or (C)-type self-deception.5 The University of Stirling Stirling FK9 4LA, UK [email protected] 4 I leave aside the problem that it seems perfectly possible to deceive oneself into believing a hypothesis for which the subjective cost of false acceptance is extraordinarily high. 5 I am very grateful for two sets of useful comments from a referee for this journal. This paper is an extended version of a piece published in Italian in a special issue of Sistemi Intelligenti (Bermúdez 1999). conservativeness and translation-dependent t-schemes 319 References Bermúdez, J. L. 1997. Defending intentionalist accounts of self-deception. Behavioral and Brain Sciences 20: 107–8. Bermúdez, J. L. 2000. Autoinganno, intenzioni e credenze contraddittorie. Un commento a Mele. Sistemi Intelligenti 3: 521–31. Haight, M. 1980. A Study of Self-Deception. Brighton: Harvester Press. Mele, A. 1997. Real self-deception. Behavioral and Brain Sciences 20: 91–102. Mele, A. 1998. Motivated belief and agency. Philosophical Psychology 11: 353–69. Nozick, R. 1981. Philosophical Explanations. Cambridge Mass.: Harvard University Press. Talbott, W. 1995. Intentional self-deception in a single, coherent self. Philosophy and Phenomenological Research 55: 27–74. Trope, Y. and A. Liberman. 1996. Social hypothesis testing: cognitive and motivational mechanisms. In Social Psychology: Handbook of Basic Principles, ed. E. Higgins and A. Kruglanski. 239–70. New York: Guildford Press. Conservativeness and translation-dependent T-schemes Jeffrey Ketland 1. The conservativeness of the (restricted) T-scheme (T) The basic principle of many popular deflationist accounts of truth1 is the disquotational T-scheme (T): 1j2 is true in L if and only if j, where to avoid semantic paradox we need a restriction: it is natural to insist that the sentence j must be a sentence of the object or base language L.2 In Ketland 1999 I reported the following theorem: 1 E.g. Paul Horwich’s minimalist account of truth (see Horwich 1998). Horwich’s theory is based on the propositional version of the disquotation scheme, where the singular term on the lhs is a ‘that’-clause rather than a quotation (Horwich calls this the ‘Equivalence Schema’). Fortunately, it makes no difference to the sort of formal investigation I am concerned with: all that matters to the formal investigations is that you have an (injective) mapping from sentences to singular terms. Terms such as ‘that snow is white’ are singular terms obtained from sentences by an injective mapping. 2 But see Theorem 4 below, where we introduce (and prove consistent) an unrestricted T-scheme, in which the truth-in-L condition for any sentence j containing ‘true-in-L’ is some arbitrary absurdity. Analysis 60.4, October 2000, pp. 319–28. © Jeffrey Ketland