2.2.1 The Theory

Theories of biological deception appeal to selection pressure, payoffs, or functions of behaviours, traits, or states (of the world) rather than the deceiver’s intentions. The most common philosophical analyses of biological deception are signalling-based and, on the predominant (‘Skyrmsian’) view, a signal S is deceptive iff (1) S is false, (2) transmitting S systematically benefits the sender, and (3) the receiver is misled or suffers harm from responding to the signal (e.g., Searcy and Nowicki, Reference Searcy and Nowicki2005; Skyrms, Reference Skyrms2010; McWhirter, Reference McWhirter2016; Shea et al., Reference Shea, Godfrey-Smith and Cao2018; Fallis and Lewis, Reference Fallis and Lewis2019, Reference Fallis and Lewis2021). Because the sender systematically benefits from sending this false signal, there is a mechanism (e.g., selection pressure) that reinforces the sending of such a signal (Fallis and Lewis, Reference Fallis and Lewis2019: 2283). Thus, it is not an accident or a mistake that the receiver is misled or harmed.

According to one rival signalling-based analysis (Birch, Reference Birch2019), deception requires that, by sending a deceptive signal, a sender strategically exploits an adaptive disposition in a receiver by raising the probability, from the receiver’s standpoint, of a non-actual state of the world. Therefore, we have deception iff (1) signal S changes subjective probabilities of states in the wrong direction, (2) sending S is a part of the sender’s strategy that exploits a particular adaptive disposition in the receiver to perform behaviour B in the non-actual state of the world whose subjective probability the signal raises, and (3) the receiver is misled by S. A strategy is any pattern of behaviour that can be transmitted by an inheritance system down the generations and characterised with conditionals of the form ‘If in context C, perform behaviour φ / activate trait Φ’ (Birch, Reference Birch2019: 31). The fact that the exploitation is strategic (involves a transmittable pattern of behaviour) and that the strategy targets particular dispositions in the sender eliminates situations in which the signal is false by accident or mistake (Birch, Reference Birch2019: 31). The sender need not benefit from this.

Finally, according to Artiga and Paternotte’s (Reference Artiga and Paternotte2018) ‘functional’ analysis, a state of the world M is deceptive iff (1) M has the (etiological) function of causing a misinformative state (or to prevent the acquisition of new information) and (2) M leads to a misinformative state. In presenting this view, I will refer to functions of behaviours (e.g., playing dead) or traits (e.g., teeth-like colouration on one’s skin) of organisms rather than states to make the argument easier to follow. The fact that the function of M is to mislead eliminates situations in which misleading was an accident or a mistake. The deceiver need not benefit from this. Also, since some strategies of deception do not seem to involve signalling (Artiga and Paternotte, Reference Artiga and Paternotte2018, Reference Artiga and Paternotte2024; Krstić, Reference Krstić2025), the functional view has a broader scope than the signalling-based analyses.Footnote ³

Because it sits comfortably with the existence of altruistic deception (Artiga and Paternotte, Reference Artiga and Paternotte2018; Birch, Reference Birch2019) and non-adaptive self-deception (§5), and because it is not limited to deception involving signalling (Artiga and Paternotte, Reference Artiga and Paternotte2018, Reference Artiga and Paternotte2024; Krstić, Reference Krstić2025), I proposed (Krstić, Reference Krstić2024c) an analysis of human deception developed along the lines of Fallis’s (Reference Fallis2015a) functional analysis of disinformation and Artiga and Paternotte’s (Reference Artiga and Paternotte2018) functional account of deception. Let us first present the theory on a simple example of biological deception and then cash it out as a general theory of human deception.

Western hog-nosed snakes deter predators by simulating death. This is deception on my version of the functional view because

1. (1) ‘Deterring predators’ (F) is the (beneficial) result of ‘the snake’s simulating death’ (M) in context C (‘predators present’),

2. (2) In C, simulating death (M) generates this result (F) by misleading,

3. (3) Misleading is the function of simulating death in C, and

4. (4) The snake’s simulating death (M) misleads the predators.

The variables are F (a specific result), M (a specific behaviour/trait), and C (the relevant context). The idea is that because the function of M in C is to mislead and M generates F in C by misleading, M systematically generates F in C. When cashed out as conceptual analysis, we get that, for a specific behaviour/trait M, result F, and context C, we have deception when

1. (1) F is the result of M in C,

2. (2) F is generated by misleading in C,

3. (3) The function of M in C is to mislead, and

4. (4) (In part) because of (3), M causally contributes to misleading.

Six important clarifications need to be made. First, I use ‘mislead’ in the meaning of ‘move further away from the truth’ or ‘cause a false representational state’ (Artiga and Paternotte, Reference Artiga and Paternotte2024). On this terminology, a misleading signal is a signal that carries misinformation, a false signal (‘misleading’ refers to its content), not a signal designed to mislead. A signal whose function is to mislead is deceptive. Also, some believe that a deceptive signal is always false. Fallis (Reference Fallis2015a), for example, defines disinformation as misleading information that has the function of misleading. However, on my view, some deceivers may use true signals to cause epistemic harm (Krstić, Reference Krstić2025, Reference Krstić2024b: §3, §4); thus, a deceptive signal need not be misleading (i.e., false). To avoid confusion, I will refer to a misleading signal as a ‘false signal’ or ‘misinformation’. Relatedly, I cannot define what counts as ending up misled or epistemically worse off in all situations – since what counts as epistemic harm is notoriously difficult to define – but this should not be seen as a debilitating flaw. My goal is not to provide a definition of misinformation. I just need concepts that may help me explain the nature of deception, and we do have a sufficiently good grasp of the relevant concepts.

Second, my analysis provides a minimal condition for deception, which is that the function of the thing that misleads is to mislead. This is consistent with the existence of intentional deception. If I φ intending to thereby deceive, then my intention gives us the function of my φ-ing. Say that I intend to trick you into wiring me $10 (‘for the transfer fee’) by telling you that your distant uncle from Nigeria left you a fortune. I lie to you intending to get your money, but I intend to get your money by causing you to believe a falsehood. Causing you to believe this falsehood is in the description under which my lie is intentional. I intend both to deceive you and to get your money.

Third, I understand ‘function’ and ‘result’ as defined by Garson’s (Reference Garson2019) generalised selected effects theory (GSE). According to GSE, in short, a proper function of something is whatever it was recently selected for by natural selection or some comparable selection process. Proper functions are proximal functions. The remainder is beneficial consequences or more distal selected effects of the thing performing its function; the remainder is the result. The function of the heart is to beat, not to circulate blood. The circulation of the blood is a beneficial result of the heart’s performing its function (Garson, Reference Garson2019: §7; Fagerberg and Garson, Reference Fagerberg and Garson2024; against, Artiga, Schulte, and Fresco, forthcoming). Analogously, when I lie to you about the supposed uncle from Nigeria, the function of my lie is to cause you to believe that you have a dead uncle in Nigeria. Getting your money is the beneficial result of you believing my lie, and (as said) I intend both the result and the function (to deceive).

On this theoretical framework, just as the circulation of the blood is the beneficial result of the heart’s performing its function (to beat), the deterring of the predators is the beneficial result of the snake’s simulating death performing its function (to mislead). Typically, the result explains why a specific trait or behaviour has this specific function in this specific context. The snake’s simulating death has a systematic tendency to deter predators (produce a certain effect) in the relevant context precisely because its function is to mislead, which explains why the snake engages in deception rather than in honest communication; revealing the truth will not generate this effect. However, misleading need not be directed at some specific target. Even though the snake engages in deception to deter predators (i.e., this result explains why it plays dead), it may also mislead non-predators. Misleading non-predators also counts as deception (the function of the behaviour is to mislead) even though this result does not explain why the snake plays dead.

Fourth, the distinction between ‘result’ and ‘function’ is crucial. Normally, the reason for deceiving is the deceiver’s practical benefit (the ‘result’) and misleading is the means of achieving this end: the deceiver gets what he ‘wants’ by misleading the dupe. The result, thus, explains why an organism engages in deception in this context, while the function explains why this is deception. Causing you to acquire the false belief that your distant uncle from Nigeria has left you a fortune is a means by which I get your money, and it explains why I intend to deceive you. However, my getting your money (the result) does not constitute deception – since you may give me the money knowing that I am trying to con you. Rather, it is causing you to believe my lie (the function) that constitutes deception, and if this happens, we have deception even if you do not give me the money.

My fifth important point is that satisfying conditions 3 and 4 (function) constitutes deception; these conditions are necessary and jointly sufficient for deception. Conditions 1 and 2 (result) are there to provide further information relevant to the ensuing analysis of interpersonal non-paradigmatic deception and self-deception. With more relevant similarities between examples of non-paradigmatic human deception and uncontroversial cases of biological deception, my argument will be stronger. This is why I will compare exemplary cases of biological deception with some cases of human behaviour and, based on the analogies, argue that human behaviour may involve deception even when the deceivers do not intend to mislead.

Finally, this terminology has a threefold purpose regarding the relevant analysis of self-deception. First, it easily captures both intentional and non-intentional deception. Second, it makes it clear that achieving some practical result (i.e., alleviating anxiety, increasing one’s happiness) by misleading the self is not a sign of self-deception – since misleading could have been an accident or a mistake. Thus, the terminology should keep us focused on what is essential for self-deception. Third, establishing this focus allows us to classify the relevant cases correctly. The terminology, that is, guides our inquiry in more productive ways.

Let us apply this functional analysis to an exemplary case of biological deception. This case will serve as a benchmark against which I will compare analogous cases of human deception, including self-deception.

The yellow-green flash is standardly understood as meaning something like ‘I am a Photinus female ready to mate’ or as raising the probability of the state in which this female is present (Skyrms, Reference Skyrms2010; Birch, Reference Birch2014, Reference Birch2019; Shea et al., Reference Shea, Godfrey-Smith and Cao2018; Skyrms and Barrett, Reference Skyrms and Barrett2019; Fallis and Lewis, Reference Fallis and Lewis2021). Therefore, the signal is false when sent by the predator Photuris female, and because it is false and sending it both systematically benefits the sender and is a part of a strategy that exploits the receiver’s adaptive disposition to descend to mate, the signalling-based analyses classify this as deception. This is deception in my view as well:

1. (1) The predator female transmits the yellow-green flash because the food tends to come down to her when she sends it,

2. (2) The signal generates this result (bringing her food) by misleading,

3. (3) Misleading is the behaviour’s function in this context, and

4. (4) Because males descend to mate rather than to get eaten, they count as being misled.

However, and this is vital for the correct understanding of human non-paradigmatic deception, the predator female does not send this signal because she ‘wants’ to mislead. Her behaviour is an evolutionary adaptation (a product of simple trial and error) caused by the fact that misleading rather than informing harmless males in this context causes them to descend. Simply put, the predator female sends this signal because the food tends to come down to her when she sends it. Her food comes down because her food-calling flash is their mating flash, but she does not know that. She only has the beneficial result ‘in mind’ and this result fully explains why she sends this specific signal in this specific context (i.e., when she ‘wants’ to eat) and why the signal means ‘I am a Photinus female ready to mate’ rather than something else (e.g., ‘I am blue’).

Paradigmatically, people know exactly what they are doing when they engage in deception, but I think that many cases of interpersonal deception are analogous to Fireflies: some people engage in a certain behaviour and are aware of their behaviour’s result but not of its function. This is why I understand human deception and self-deception in functional terms. The predator ‘knows’ that the food tends to come down when she sends the signal, but she does not ‘know’ that the food comes down because the signal misleads it. Analogously, because they are unaware of their behaviour’s (trait’s) function, non-paradigmatic human deceivers do not intend to mislead even when they intend the result. The behaviour’s function is an adaptation (cultural, evolutionary, etc.) caused by the fact that a specific kind of dishonesty systematically generates the required response on the part of the victims. Intentional deception is caused by the same fact; the only difference is that intentional deceivers also intend the means whereby they generate this response.

This might seem complicated, but it is actually very simple. To show what I mean, I proceed to compare cases of non-paradigmatic human deception with Fireflies (§2.2.2); the analogies between them are striking. I will then consider an objection to my analysis (§2.2.3) and proceed to develop the relevant functional analysis of self-deception (§3).

2.2.2 Non-Paradigmatic Human Deception

Recall, my view easily applies to intentional deception. When I tell you that your distant dead uncle from Nigeria has left you a fortune and that I need $10 for processing fees, I intend to get your money by causing you to believe my lie. Since proper functions are proximal functions, the function of my lie is to mislead, and it is a part of the description under which my behaviour is intentional. Getting your money is the lie’s beneficial result, just as the circulation of blood is the beneficial result of the heart’s beating.

Let us now apply the view to non-intentional interpersonal deception. Consider an interesting piece of advice for APA interviews (Krstić, Reference Krstić2021, Reference Krstić2024c). The advice is Don’t ‘be yourself’ (Arvan, Reference Arvan2015). Rather than being yourself, you should create

People who have just finished their PhD can hardly have a steely-eyed grasp of the real needs of actual hiring departments, they are rarely confident in their teaching and research, and they often apply for jobs for which they are not a perfect fit. For these job seekers, the advice ‘don’t be yourself’ genuinely means ‘present yourself as a different person, the perfect candidate’. This advice is sensible, but this is also exactly what one predator female firefly would say to another. Just as a job candidate does not want to transmit a message that means ‘I am not the best candidate, but I need you to give me this job’, the predator firefly does not ‘want’ to transmit a signal that means ‘I eat harmless males and I need you to come down’. The ‘not being themselves’ job candidates pretend to be the perfect candidate by mimicking the behaviour of the perfect candidate in the same way the predator female firefly pretends to be a sexually receptive female by mimicking her signal.

In most cases, job candidates do not realise that they are acting deceptively by not being themselves, and, thus, they do not intend to deceive. And even though they only intend to get the job by not being themselves, that is, they intend the behaviour’s result, misleading is nonetheless the function of their behaviour: the members of the search committee should think that they are the perfect candidate. An extremely high selection pressure explains the function of this behaviour: because only 3.5 per cent PhDs manage to secure a position in academia (Taylor et al., Reference Taylor, Martin and Wilsdon2010: 14), candidates have the best chances of getting a job by presenting themselves as being perfect for the job. Therefore, when this behaviour misleads, this is not an accident or a mistake. Not being oneself in this context is deception because

1. (1) Getting the job is the beneficial result (F) of this behaviour (M) in this context (C),

2. (2) This result is generated by misleading,

3. (3) The function of the behaviour in C is to mislead, and

4. (4) The behaviour misleads.

One reason to think that this is not deception is that conditions 3 and 4 are not satisfied: everyone understands that this is a formal interview and that a candidate’s behaviour is a performance, like being on stage. One could say that, if the members of the search committee are impressed by the candidate and think that they are clearly hireable, that would count as a success of the candidate’s performance – since they are trying to make a good impression in the interview. And because everyone sees this as performance, the function of the behaviour is not to mislead, and it does not mislead. I think differently. We cannot compare academic job interviews with performances of this kind. The candidate’s performance must be an indicator of actual qualities: Havier Bardem did not get an Oscar for his role in ‘No Country for Old Men’ because the committee thought that he would be a great psychopathic killer. Moreover, unlike performances of figure skaters or gymnasts, the ‘not being themselves’ candidates mimic qualities they do not have. Therefore, even if everyone really thinks that this is a formal interview and that a job candidate is ‘performing’, unlike a figure skater but just like an actor, a candidate can purposefully mislead by mimicking qualities that they do not have. For instance, they could mispresent themself as having exceptional tolerance or patience.

Let us now consider a case that involves using the poisonous ‘belladonna’ (Atropa belladonna) plant to make one’s pupils dilate and acquire a dusky, lustrous appearance of one’s eyes. Because this appearance was the height of beauty in Renaissance Italy, Venetian ladies used belladonna extract to dilate their pupils and make themselves look more beautiful (Passos and Mironidou-Tzouveleki, Reference Passos, Mironidou-Tzouveleki and Preedy2016: 766; Carlini and Maia, Reference Carlini, Maia, Gopalakrishnakone, Carlini and Ligabue-Braun2017: 66). Dilated pupils make people look more beautiful for a very important reason: people’s pupils dilate the most when they are looking at someone they find sexually stimulating. Men and homosexual women are most attracted by large pupils in women, while heterosexual women are most attracted by medium-sized pupils in men (Tombs and Silverman, Reference Tombs and Silverman2004; Rieger and Savin-Williams, Reference Rieger and Savin-Williams2012) and, correspondingly, women’s pupils dilate more than men’s (Lick, Cortland, and Johnson, Reference Lick, Cortland and Johnson2016). Arguably, this discrepancy in preferences exists because males and homosexual females are most attracted to large pupils because their reproductive strategies are best served if they can avoid missing a mating opportunity with an interested partner (Tombs and Silverman, Reference Tombs and Silverman2004; Lick, Cortland, and Johnson, Reference Lick, Cortland and Johnson2016). Heterosexual females are predisposed to favour more moderate sexual attentions to avoid possible physical harm.

When we put this together, we gain two important insights: First, the false impression that a woman with large pupils is sexually aroused makes her appear more attractive; people think that the woman looks more attractive because they are misled by the size of her pupils. Second, generating this false impression is the function of dilating pupils. The parallel with Fireflies is clear: the woman sends a false signal about her sexual receptiveness by dilating her pupils just as the predator female firefly sends a false signal about her identity and sexual receptiveness by sending a yellow-green flash. Therefore, using belladonna to send a false signal in Renaissance Venice involves deception for the same reasons the predator female firefly’s sending the false signal involves deception. On my analysis, using belladonna involves deception because:

1. (1) The result of dilating pupils (M) in this context (C) is to look more attractive (F).

2. (2) This result is in C generated by misleading: a woman looks more attractive because she gives the false impression that she is sexually aroused.

3. (3) The function of dilating pupils in C is to mislead (cause this false impression).

4. (4) Dilating pupils misleads.

The most important similarity between the Belladonna and Fireflies is this. Because Renaissance people did not have access to the relevant studies, those who used belladonna could not have intended to give this false impression. They thought that what makes them more attractive are larger pupils qua larger pupils in the sense in which blue eyes qua blue eyes may make a person pretty. They knew that they would look prettier by dilating pupils, but they did not know why. Therefore, they intended only the result (looking more attractive). Not realising that you are being deceptive by behaving in a certain way is important because you cannot intend to generate X by φ-ing if you do not know that φ-ing generates X. If a person does not know that using belladonna causes a misleading impression, that is, she cannot intend or expect to cause it (compare, Krstić, Reference Krstić2020b, Reference Krstić2023b) and, thus, she cannot intend to mislead. Therefore, these women engaged in deception, but they did not intend to mislead. They thought that their dilated pupils were a sign of beauty just as the predator Photuris fireflies ‘think’ that their signal means ‘Food, come down’.

Let us analyse Belladonna by considering a passage from Davidson (Reference Davidson1963: 686–687), who writes [italics and a comment added]:

What rationalises dilating pupils is that the Renaissance women want to look more attractive (they intend the result). Just as Davidson flips the switch to illuminate the room, they use belladonna to look more attractive. However, flipping the switch does not directly illuminate the room; rather, one illuminates the room by turning on the light. Likewise, unbeknownst to everyone, dilating pupils does not directly make a person look more attractive; rather, people with dilated pupils look more attractive by causing the relevant misleading impression. And just as the function of turning on the light is to illuminate the room, the function of using belladonna is to cause this misleading impression. Therefore, we have deception when dilating pupils misleads, just as we illuminate the room when we flip the switch.

However, unlike Davidson, who knows that he will illuminate the room by turning on the light, which is why he intends to turn on the light, and unlike myself who expects to get your money by causing you to falsely believe that your dead Nigerian uncle left you a fortune, Renaissance women did not know that they would look prettier by giving this false impression. Rather, they thought that flipping the switch (dilating pupils) directly illuminated the room (made them look prettier) without the ‘mediation’ of turning on the light (misleading). Therefore, they intended only to illuminate the room (look prettier) by flipping the switch (dilating pupils); they did not intend to deceive (turn on the light).

Similar behaviour occurs in contemporary society as well. Some people put on makeup so subtly that it looks like they are not wearing makeup at all. Because the function of this behaviour typically is to cause false beliefs about one’s appearance and age, they are engaged in deception, but we should not think that they all intend to mislead. Most of them simply do this because it is fashionable, because they want to look younger, and so on. They do not intend or expect to mislead.

Before I proceed, I need to resolve some objections to my argument.

2.2.3 Resolving Some Concerns

The way to know whether the function of behaviours of people putting on makeup or ‘not being themselves’ job candidates is to mislead or something like ‘securing my practical interests’ is to identify the behaviour’s proximal function in the relevant context. If securing my practical interests is the proximal function of my behaviour or trait, then I do not count as engaging in deception and the epistemic harm was caused by an accident or a mistake. However, if the practical interests are being secured by misleading, then misleading is the function of my behaviour (it is the most proximal activity in the sequence) and securing these practical interests is its result.

One worry may be that my main argument is viciously circular: I use cases of supposed non-paradigmatic deception to support my analysis of human deception, but I then use my analysis to argue that these cases involve deception. Therefore, I will show that theories of biological deception also classify my cases as involving deception and that each case has its clear analogue in animal deception.

Skyrmsian analysis counts my cases as deception: the signals are false, the senders systematically benefit from sending them, and the receivers suffer harm and are misled. Even though job interviews are performances, they are the kind of performances that should be reliable indicators of the candidates’ real qualities. Therefore, job candidates can transmit false signals and these signals systematically bring benefit to some of them when the committee is misled or harmed (by hiring a ‘not-perfect’ candidate). Also, when people put on makeup or dilate pupils, they transmit false signals regarding their appearance and these signals systematically bring them benefit when they mislead or harm the receivers. Therefore, this is all deception.

Birch’s strategic exploitation view also counts them as deception. According to this view (Birch, Reference Birch2019: 31), a signal S, sent in a state of the world X, is biologically deceptive iff:

1. (a) Sending S in X elicits some behaviour B in the receiver.

2. (b) S elicits B in X not because B benefits the receiver in X, but because (i) B benefits receivers in some other state of the world, X’, and (ii) P(X’|S) > P(X’).

3. (c) S is sent in X as part of a strategy.

4. (d) The sender’s strategy has been maintained by selection, at least in part, because of the payoffs conferred by the receiver’s performance of B in X.

In this state of the world X, at least for some members of the search committee, the candidate’s behaviour raises the subjective probability of a non-actual state of the world X’ in which the candidate is perfect for the job, and search committees hire candidates (behaviour B) when they are in X’. This behaviour is part of a strategy that has been maintained by selection, at least in part, because of the payoffs conferred by search committee members’ performance of B in X: candidates present themselves as the perfect candidate because this is who search committees hire. Therefore, this is deception.

Similarly, in this state of the world X, dilating pupils raises the subjective probability of a non-actual state of the world X’ in which the person is more attractive, and people tend to look (more) favourably (behaviour B) on people when they are in X’. These behaviours involve a strategy maintained by selection, at least in part, because of the payoffs conferred by other people’s performance of B in X: people improve their appearance because this allows them to leave a (more) positive visual impression on others (the payoff). Therefore, this is deception.

These examples also have clear analogues in biological deception. In a commonly discussed kind of automimicry, some members of a species resemble their better-defended conspecifics. One example is the Monarch butterfly (Danaus Plexippus): some monarchs do not contain toxic cardiac glycosides, but predators cannot easily know which ones do. This is how job candidates mimic perfect candidates. Similarly, some animals have colouration on their head that makes their horns and antlers look bigger or their ears are positioned next to horns to make it look like the animal has more horns (Guthrie and Petocz, Reference Guthrie and Petocz1970; West-Eberhard, Reference West-Eberhard1979: 226) and some plants give an impression of more extensive thorns by having colourful elongated organs found in some plants (Lev-Yadun, Reference Lev-Yadun2003). These organisms are not misrepresenting themselves as someone else. Rather, the function of colouration, ears, or fake thorns is to exaggerate the already existing qualities: they make the horns look bigger or give the impression that the organism has more horns or thorns. Using the belladonna extract or putting on makeup is analogous to these cases: their function is to exaggerate an individual’s existing qualities.

In conclusion, these cases involve deception on other theories of biological deception, and they have clear analogues in animal deception. Therefore, intending to deceive is not necessary for interpersonal deception. The time has come to present the functional analysis of self-deception.

3.2.1 Introduction

Because not all self-deceivers deceive themselves by behaving in obviously deceptive ways, it would be natural to spell out some deceptive behaviours and traits. This is something I cannot do, but for a very good reason: context partially determines whether a certain behaviour or trait is deceptive (§2.2.1, §3.1) and, thus, what might be an honest behaviour in some contexts will have a deceptive character in others. Because, in the right circumstances, pretty much any trait or behaviour can be used to deceive – even sending true messages (Krstić and Saville, Reference Krstić and Saville2019; Krstić, Reference Krstić2024ca, Reference Krstić2025) – and because I model self-deception on interpersonal deception, pretty much any behaviour or trait can be self-deceptive in my view. So, how to identify self-deception, then? This section tries to answer this question.

Philosophers predominantly think that self-deceivers cause themselves to acquire or retain incorrect and epistemically unjustified beliefs (credence in propositions) and, on the most popular view, we can identify self-deceivers by comparing their beliefs with beliefs their impartial cognitive peers, ICPs, would form (Mele, Reference Mele2001: 106–107, Reference Mele and Hirstein2009: 146–147; Lynch, Reference Lynch2012; Dolcini, Reference Dolcini, Felline, Ledda, Paoli and Rossanese2016; similarly, Galeotti, Reference Galeotti2018: 45). If most of the ICPs would examine the same evidence using as much reflection on the same issue as her and form a different belief, the person is deceiving herself.

I think that we should instead compare how a person normally reasons with how she reasons now. If we have reasons to think that, due to some motivating factor, she does not behave the way she normally behaves, then we should ascribe self-deception to her even if we or her ICPs think that her belief is epistemically justified. Self-deception is characterised by a kind of motivated internal irrationality: the person is being inconsistent; she is not behaving as she normally does. And because this kind of irrationality can be identified from the outside – namely, we can notice that self-deceivers do not behave the way they normally behave – we can identify self-deception by seeing whether the person is not being herself.

An immediate worry with this proposal is the issue of whether we have a grasp of how an individual typically behaves sufficient to identify self-deception by noting the departure from this behaviour. We often have no experience or history with those to whom we attribute self-deception. I think that we should not ascribe self-deception to those whom we do not know well enough. More importantly, if we have no experience or history with the person, then we do not know how she typically reasons and, thus, we cannot identify her genuine ICPs. This is why most scholars take the ICP criterion rather loosely. Instead of looking for genuine ICPs, we use ourselves as a model of epistemic ‘rectitude’: given the evidence, here is what (we think) should be believed. And the notable difference is interest or desire: but for the motivational states, the individual is not likely to have come to believe what they believe; rather, they would have believed what I think they should believe. I fear that this is a recipe for false positives. And even if we try to find the person’s ICPs is sometimes rather dangerous. Consider the case of Ignaz Semmelweis.

Notice, it is perfectly natural to think that the members of the relevant medical community were Ignaz’s ICPs. True, some of them were offended and thus perhaps not impartial, but their view had overwhelming evidential and theoretic support at that time. It was just that the evidence was misleading and the theories used to explain it were false. On the view that the belief Ignaz’s ICPs would form is the benchmark against which we would judge whether Ignaz is a self-deceiver (i.e., whether he exhibits the appropriate motivation), we get a clear false positive here. Therefore, we have a good reason not to rely on the ICP test in such a context: it may misclassify people who jump out of the system as being self-deceived.

The idea that self-deception involves motivated internal irrationality is not novel. Davidson (Reference Davidson1986) famously insists that self-deception involves a person violating her own norms. Similarly, Szabados (Reference Szabados1974) mentions that self-deception involves quasi-rationality. Michel and Newen (Reference Michel and Newen2010) suggest that self-deception involves applying double standards in evaluating beliefs relevant to one’s self-esteem: the person’s belief-evaluation method depends on the anticipated result rather than on the assessment of the method itself. Finally, on Mele’s account (Reference Mele1987, Reference Mele1997, Reference Mele2001, Reference 68Mele2020), because they treat (seemingly) relevant data in a motivationally biased way, self-deceivers do not reason the way they normally reason.

Because we are not being our regular selves when we reason in a biased way, it may seem that I am merely reformulating a widely shared hypothesis. However, this appearance is misleading. I also argue that self-deceivers need not end up with epistemically unjustified mental states, their reasoning need not be normatively faulty, and it need not deviate from the reasoning of their ICPs. The bias can even improve the self-deceiver’s reasoning by trumping a systematically exhibited reasoning failure. This is something most philosophers straightforwardly deny. For example, Barnes (Reference Barnes1997: 78) writes that ‘as a result of this bias in the self-deceiver’s belief-acquisition process, the self-deceiver’s belief that p is never justified[;] [the] self-deceiver will never have adequate grounds for thinking that his or her belief that p is true’ (similarly, Van Leeuwen, Reference Van Leeuwen2009: 112; Dolcini, Reference Dolcini, Felline, Ledda, Paoli and Rossanese2016; Galeotti, Reference Galeotti2018: 52).

I am not advocating for a radical break-up with the idea that self-deceptive beliefs are epistemically problematic. I just want to modify it. Even when the self-deceiver’s belief is epistemically unjustified, it is unjustified because the person does not behave how she normally behaves. Also, I do not deny that self-deception is a failure to believe in accordance with one’s evidence. I just argue that what counts as one’s evidence is not for us or one’s ICPs to decide. One’s evidence should be determined based on how one normally reasons. When people come to see that they have self-deceived themselves and regret it, what they regret is that they had failed to mind the evidence. And they regret not being themselves because they take themselves as possessing the competence to respond to evidence appropriately. Importantly, they may think that they possess this competence even if everyone else thinks differently; this was surely true of Ignaz.

I now proceed to present my theoretical apparatus and then argue that my view should be preferred.

3.2.2 Not Being Yourself

Davidson (Reference Davidson1985) famously distinguishes when a mental state or action is irrational or incoherent from the third-person perspective, which is when a person has a set of attitudes that are inconsistent with someone else’s standards, from when it is irrational from the first-person perspective, which is when her attitudes are inconsistent by her own standards. Let us call this first-person kind of irrationality (inconsistency) internal.

I argue that internal irrationality characterises self-deception even though self-deceivers need not realise that they are misleading themselves. They may, nonetheless, notice that they are not behaving the way they would normally behave, and when they do notice this, they will exhibit some discomfort, doxastic dissonance, or behavioural tension, they may avoid some evidence, and so on. However, the cause of this discomfort or tension is not the realisation that they simultaneously believe that p and that not-p, but rather the feeling that they are doing something they typically would not do. I do not argue that no self-deceiver ‘knows the truth deep down’. Rather, I argue that, for the tension, dissonance, or discomfit to arise, it is sufficient that the person thinks, suspects, or ‘feels inside’ that she is not really being her normal self. This would also make her avoid some evidence since the evidence will remind her of her own inconsistency. Let us now draw three uncontroversial insights from the proposed distinction between external and internal irrationality.

First, a ‘third person’ can reliably judge that a mental state or action is first-person irrational: you can judge that my mental state or behaviour is internally irrational even when I fail to notice this. To identify internal irrationality from the outside, we need to focus on how the person normally reasons as opposed to how she reasons now, rather than on whether she is violating some epistemic norms or laws of logic (i.e., whether she is externally irrational). ‘Is she being herself?’ we should ask. If we suspect that the person is not really herself, an outward sign of internal irrationality, we should suspect that she is deceiving herself. The reason for this approach lies in the second insight.

Second, in certain situations, my mental state or action can be rational from the third-person perspective (external rationality) while being irrational from the first-person perspective (internal irrationality). I call this mental state or action pseudo-rational since it is normatively rational but subjectively irrational. It is normatively rational because it satisfies the normative standards of reasoning: people from my environment reason or behave in such a way. And it is subjectively irrational because those are not my standards: I am ‘not being myself’ (I am being internally incoherent) even if I do not notice it. Notice, when I say that those are not ‘my’ standards, I do not mean that a person who, for example, systematically jumps to conclusions endorses jumping to conclusions. Rather, she jumps to conclusions thinking that this is the correct way to reason in the given context. She endorses this behaviour not knowing that her reasoning is erroneous.

Third, the self-deceiver’s ‘not being oneself’ is analogous to the ‘not being oneself’ of the Belladonna women or the predator female firefly. The predator female firefly does what she normally does when she ‘wants’ to eat; she is being regular herself in this sense. However, she is not being regular herself in the sense that she is misrepresenting herself as someone else, namely, as a sexually receptive (‘honest’) female. By analogy, self-deceivers are being their normal selves in the sense that they may standardly engage in self-deception in certain contexts, but they are not being their normal selves in the sense that this is not how they would normally behave had they not been motivated to believe a certain proposition. Because the ‘not being oneself’ of the self-deceiver preserves the analogy between self-deception and other-deception and since this kind of irrationality can be detected from the outside, this is the best way to identify self-deception. I now proceed to defend this position.

3.2.3 The Argument

According to the most popular view, the way to identify the level of emotional or motivational bias appropriate for a person self-deceived about something – that is, the way to identify a self-deceiver – is to see what kind of belief the person’s ICPs would acquire or retain. If the majority of her ICPs would examine the same evidence using as much reflection on the same issue as her, and conclude that her belief is false, we can say that the person is a self-deceiver. Impartial cognitive peers are people very similar to the person being tested in things such as education and intelligence, but who do not have an emotional stake in the issue at hand, who do not desire any relevant proposition to be true, and who would not prefer any evidence over some other evidence (Mele, Reference Mele and Hirstein2009: 146–147).

I argue that focusing on whether the person is not being herself is a better way to identify self-deception. My argument involves discussing five cases, starting with Ben.

C is a context in which Ben makes inferences to the best explanation regarding people’s behaviour. Because Ben systematically jumps to conclusions in C, we get that Ben systematically makes faulty inferences to the best explanation when interpreting other people’s behaviour, and his relevant beliefs are systematically unjustified. However, this time, C involves assessing the behaviour of Ben’s partner concerning a very sensitive topic. If Ben were to jump to a conclusion, he would infer that she is having an affair. However, because he desperately wants it to be true that his partner is not having an affair, Ben exhibits a reasoning bias. And the bias cannot simply make Ben’s poor reasoning worse, since this would only make him more confident in the proposition he does not want to be true. Rather, because jumping to conclusions in C leads to a proposition Ben does not want to be true, the bias must improve his reasoning so that he now carefully evaluates all the evidence and avoids making the conclusion he does not want to be true. Ben’s reasoning now is pseudo-rational: it is normatively rational but internally irrational. And because they do not jump to conclusions in C, Ben’s friends form the same belief as Ben.

One might object that because Ben’s reasoning appears to differ systematically from that of other humans, these humans are not his ICPs. Ben’s real ICPs would commit F in C and would, thus, not share Ben’s views. Therefore, the ICP test succeeds; we just need to find Ben’s real ICPs. I am not sure that ICPs must think just like unbiased Ben, but instead of defending this view, I ask: if Ben’s IPCs need to think just like unbiased Ben, why do we need them? We just need to consider what Ben would normally conclude in this context, had his bias been absent. We need to know this if we want to find his ICPs anyway. So, seeing that Ben is not being his regular self comes before identifying his ICPs.

The ‘not being yourself’ hypothesis also preserves a nice analogy with other-deception. Ben acts like a person who systematically does not jump to conclusions because this satisfies his desire, just as the predator female firefly acts like an ‘honest’ female because this brings her food. To get what they ‘want’, both deceivers must not be themselves: Ben will not believe the welcome proposition if he continues to jump to conclusions, and males will not descend if the predator sends a signal that means ‘I am hungry and I need you to come down’. Importantly, Ben’s false belief is luckily epistemically justified and consistent with the belief that his friends also formed. Because Ben now considers all his evidence, the assessment of his environment cannot tell us whether he deceived himself in C or not. Therefore, analyses based on the idea that self-deceivers end up with unjustified beliefs will not classify this as self-deception.

My functional view classifies this as self-deception. Not only that Ben’s belief that not-p is false, he is also misled about which standard of reasoning he would normally use in C and about the reason why he is deviating from his typical reasoning. Therefore, Ben is deceived about whether p – condition 4. Had Ben reasoned the way he usually does, he would reach p (affair!) rather than not-p; he is ‘not being himself’ – condition 3. And just as the predator female firefly can lure the honest males only by not being herself, Ben can end up believing that not-p only by not jumping to conclusion, which explains why he ‘is not being himself’ and gives us good reasons to say that the function of his behaviour is to mislead – condition 2. Ben’s ‘not being himself’ is a one-person analogue of the behaviour of women using belladonna or predator female fireflies; they are all not being themselves. Satisfying Ben’s desire that his partner is not having an affair (condition 1) is the result of this behaviour, and this result explains why Ben engages in self-deception.

One may argue that even if Ben is propositionally justified in believing that his partner is not having an affair (he has good reasons to believe this), he is not doxastically justified in so believing (he does not form his belief based on those reasons) – since he is self-deceived in coming to believe that his partner is not having an affair. Therefore, the belief is not sufficiently justified, and the idea that self-deceivers always fall below some epistemic standard is salvaged. However, the fact that Ben is self-deceived in coming to believe that not-p is consistent with the fact that he forms his false belief from good reasons. The description of the case is such that the bias shifts Ben’s reasoning towards good reasons; the bias causally contributes towards the belief being doxastically justified. To this, one may respond that if Ben is moved to believe based on a normatively apt grasp of epistemic considerations, he is not self-deceived. This does not follow for two reasons.

First, Ben does not recognise the evidential goodness of the data he assesses. He assesses it in this particular way because he wants that not-p is true. If he were to reason the way he reasons when he is not biased, he would treat the same data differently: he would jump to a conclusion and believe that p. The bias involved in this case is the same kind of bias we see in paradigmatic cases of self-deception; the only difference is that the bias applies to a person who systematically reasons erroneously. Second, the fact that the belief is based on good reasons does not suggest that this is not (self-)deception: Romeo’s belief that Juliet was dead was also justified, both doxastically and propositionally. In trying to understand the cases from this section, one should think outside the standard ‘self-deception’ box and within the ‘interpersonal deception’ box. This standard ‘self-deception’ box is too small and often confusing.

Consider now the cases of Mario, his twin brother Dario, and their ICPs.

Because they commit F in C but Mario does not, Mario’s ICPs would say that he deceived himself into believing that not-p because his belief is epistemically unjustified and false. However, in reality, Mario acquired a belief that is epistemically justified but unluckily false, whereas his ICPs acquired a belief that is epistemically unjustified and luckily true. Relying on the judgment of Mario’s ICPs causes us to ascribe self-deception to Mario for the wrong reasons. Mario is a self-deceiver for the following reasons. First, his belief that p is false and he is misled about propositions relevant to whether p – condition 4. Second, he believes that not-p (no affair) due to the influence of his desire that his partner is not having an affair – condition 1. Third, had he reasoned the way he normally reasons, he would have ended up believing the opposite proposition: Mario is not being himself – condition 3. Fourth, because he is not being himself (the analogy with interpersonal deception), we can plausibly assume that the function of Mario’s behaviour is to mislead – condition 2. Thus, this is self-deception in my view.

Consider now Mario’s twin brother.

Again, because they commit F in C but Dario does not, his ICPs would say that Dario is a self-deceiver. However, they would think that Dario is a self-deceiver because his belief is epistemically unjustified and false, but, in fact, their belief is epistemically unjustified and false. Rather than focusing on his supposed ICPs or what we think Dario should believe, we should ascribe self-deception to Dario for the following reasons. First, Dario exhibits internal irrationality, condition 3, and his irrationality is motivated – condition 1. Had Dario reasoned the way he usually does, he would reach p (affair) rather than not-p. Second, because he is not being himself and because only not being himself will satisfy his desire, we can plausibly assume that the function of his behaviour is to mislead – condition 2. Finally, just like Ben and Mario, Dario is misled about which standard of reasoning he would normally use in this context and about the reason why he is not behaving the way he normally behaves; he is deceived about whether p – condition 4. Therefore, Dario is a self-deceiver.

My final two cases are of two twin sisters, Anna and Hanna.

Anna’s friends would think that Anna reasoned correctly, but, in fact, she does seem to have deceived herself about her partner’s fidelity. Her belief is false and, because she is ‘not being herself’, we have reasons to believe that she did not mislead herself by mistake or accident; this seems to be the function of her behaviour. If we focus on what we or her friends think Anna should believe in this context, we will not understand her behaviour. However, if we compare how she normally reasons in C with how she reasons now, we will understand everything. Seeing that she is not using the reasoning method she standardly uses, we will know that something is not right; Anna is not being herself. And since she has a good reason for this ‘not being herself’, we can infer that this reasoning was not an accident or a mistake, which suggests that she is deceiving herself.

Hanna’s case is not much different.

Even though, unlike her twin Anna, Hanna ends up with a justified true belief that not-p, we have excellent reasons to think that she is a self-deceiver. Just like Dario, she is deceived about whether p and she is not being herself. And she is not herself because her desire that a specific proposition is true guides her behaviour. Thus, we have a good reason to think that she did not mislead herself by accident or by mistake. However, because her belief is justified, relying on what we think she should believe gives us a false negative. Thus, in assessing self-deception, we should focus on how the person now behaves as opposed to how she normally behaves.

I end this section by restating my main point. People often engage in fallacious reasoning: they employ rationalisations or reappraisals to justify their views, they doubt the source of threatening information, they shift blame to others or try to explain threatening information or doubts away, they appraise the threatening evidence as ambiguous or support their conclusions with empirical claims that are difficult to verify or falsify, and many more. Engaging in such behaviour is not necessarily self-deceptive since this could just be how they normally behave. However, when the function of such behaviour is to mislead, namely, when this behaviour signals that the person is not being herself, then we can say that she is engaging in self-deception.

More should be said about the cases discussed here, but I will have to do this in follow-up discussions. I now proceed to apply my analysis to Maria and Sid.

4.2.1 The Received Views

The problem unwelcome self-deception poses for the standard view is twofold: not only that the view needs to identify the benefit people systematically get by making themselves believe something they do not want to be true, this benefit needs to be consistent with the benefit people systematically get from believing what they want to be true. One solution is that engaging in self-deception (welcome and unwelcome) is set off by the need to reduce anxiety (Johnston, Reference Johnston, McLaughlin and Rorty1988; Barnes, Reference Barnes1997). Barnes argues that, while welcome self-deception reduces anxiety with respect to the acquired belief, deceiving oneself into believing something one does not welcome reduces anxiety associated with some other proposition. The following case should illustrate her point.

Barnes’s explanation is that, by misinterpreting evidence, John trades higher anxiety, which is caused by believing the proposition supported by George’s behaviour, for lesser anxiety, caused by acquiring the self-deceptive belief. The more powerful desire causes John to deceive himself. However, believing the unwelcome proposition would reduce anxiety only if John cared more about George’s regard than Mary’s fidelity, which is unlikely. The belief that his wife is unfaithful actually increases John’s anxiety (Scott-Kakures, Reference Scott-Kakures2001: 321). And the hypothesis deals even worse with Zelda because her anxiety is generated exclusively by her self-deception. Therefore, reducing anxiety seems to be an unlikely general cause of self-deception.

Because unwelcome self-deception often causes or increases anxiety, some scholars suggest that what explains its existence is that it brings practical benefit at the cost of inflicting some typically short-term (side-effect) harm, such as increased anxiety, when the benefit outweighs the harm. The basis for this view was provided by Mele (Reference Mele1999: 122, Reference Mele2001: §V, Reference Mele2006: 114), who writes that, whereas for many people it is more important to avoid acquiring the false belief that their spouses are having affairs, the converse may well be true of some insecure, jealous people. Mele’s proposal is very convincing and influential, and it comes in many variations. This version is based on two theses.

The first is taken from Sharpsteen and Kirkpatrick (Reference Sharpsteen and Kirkpatrick1997: 627), who write that the jealousy complex is a mechanism for maintaining close relationships that appears to be ‘triggered by separation, or the threat of separation, from attachment figures’. Let this be the threat thesis. The second is the thesis that the concern for minimising having costly false beliefs drives lay hypothesis testing, the so-called FTL theory. Recall, according to FTL (§3.1), the greater the predicted damage caused by falsely accepting that p, the greater the cost of falsely accepting it and, as a result, the greater the aversion towards falsely believing it. And the level of aversion towards falsely believing a proposition determines the threshold for accepting or rejecting it: the lower the threshold, the less evidence is sufficient for accepting a proposition, and the threshold is low if the expected cost of believing it falsely is low (Mele, Reference Mele1999, Reference Mele2001, Reference Mele2006; see, Scott-Kakures, Reference Scott-Kakures2000).

On this view, Zelda’s strong, jealous desire to maintain her relationship with Scott plays a role in rendering the false belief that Scott is not having an affair a ‘costly’ error. Since believing erroneously that Scott is not having an affair may cause her not to take steps to protect the relationship against an intruder, Zelda would have a higher aversion towards falsely believing this proposition than towards falsely believing the one that causes her suffering (Mele, Reference Mele2001: §5, Reference Mele2006: 114; Lauria, Preissmann, and Clément, Reference Lauria, Preissmann and Clément2016: 127, n. 10; Gadsby, Reference Gadsby2022: 252). Therefore, she deceives herself because this will cause herself to take steps to protect her relationship (her benefit), and this is more important than the harm she will also bring about.

This proposal can even explain the fact that, by being unreasonably jealous, the self-deceiver may jeopardise or harm her marriage. Recall, self-deceivers are acting according to their perceived (rather than real) primary error (Friedrich, Reference Friedrich1993; see, Scott-Kakures, Reference Scott-Kakures2000) and what this error is in any particular case depends upon a person’s current motivational, affective, and attentional condition. Because costs are subjective, accepting the ‘affair’ hypothesis need not be a costly error from the self-deceiver’s perspective. We should also expect some degree of myopia in these cases that will affect the cost-benefit assessment: Zelda and John seem to be discounting the possibility of paying some distant costs, such as harming their relationships with their jealousy.

One concern is that the threat thesis limits the application of the view only to cases of unwelcome self-deception that involve jealousy (Lazar, Reference Lazar1999: 275), but this concern can be resolved. The self-deceiver need not be jealous; something else could be under threat. For instance, a view of oneself as being the sort of person who would never be taken unawares by a cheating spouse or of oneself as being very attractive or highly regarded by everyone could be under threat. Nonetheless, one problem remains: Where did Zelda get the idea that she could be taken unawares by her cheating spouse or that Scott thinks that she is not attractive?

One of the main ideas of this proposal is that the function of believing that Scott is having an affair is to help Zelda protect the relationship against the intruder; however, there is no threat of separation that could trigger jealousy, suspicion, or this kind of evidence assessment. In fact, it is the other way around: the way she assesses evidence and her distrust cause Zelda to think that there is a threat of separation. Zelda invents evidence, and she interprets counterevidence to p as evidence that p: she thinks that cheating fits Scott because he is such an unlikely person to do it. Thus, while it may be correct that Zelda’s jealousy (distrust) makes the ‘no-affair’ hypothesis a costly error, Mele’s theory cannot tell us where the threat came from. Scott’s behaviour involves nothing extraordinary, nothing that should cause her to think that she is not very attractive or that she should not trust him. The jealousy/distrust/etc. makes the ‘no-affair’ hypothesis a costly error. This explanation, thus, raises two concerns: first, it understands the aetiology of unwelcome self-deception backwards and, second, it cannot tell us what triggers self-deception in people like Zelda.

One may object that Zelda is not inventing evidence but rather merely interpreting a fact (the absence of evidence for p), or even evidence in favour of not-p, as evidence for p. Perhaps, there is a sense in which Zelda does not invent evidence when she analyses Scott’s going to the bathroom but rather puts enormous weight on what is really very weak evidence. However, this still raises the first concern: it is the jealousy (distrust, protecting self-esteem, etc.) that explains Zelda’s way of interpreting the evidence, not the other way around. Given the prior probability and the likelihood, the formation of the thought ‘Scott went to see Ernest when he was in the bathroom’ involves a very large deviation from the normal Bayesian updating. Therefore, to handle the case, the FTL view requires Zelda to have an unreasonably low threshold for what counts as evidence in support of the ‘affair’ proposition if her accepting this proposition is to generate jealousy, suspicion, or threaten Zelda’s self-image or self-respect. However, as a consequence, the first factor (jealousy, distrust, self-image) becomes redundant in the aetiology of self-deception: it cannot tell us why Zelda deceives herself.

Inventing evidence is an important feature of some cases of self-deception that cannot be explained solely by appealing to Mele’s two theses. Recall the case of parents who deceived themselves into believing that they were to be blamed for the death of their child who died of leukaemia – call it Heartbroken Parents. No information could function as evidence in support of their belief. There must be something in the parents’ or Zelda’s psychology that overrides their normal hypothesis-testing mechanisms, but the threat thesis cannot tell us what.

It could be that it is not the job of Mele’s theory to tell us where Zelda’s threat came from; this is part of the backstory rather than the theory. Maybe Zelda is standardly jealous or distrustful because she was badly hurt by the affairs of her previous partners. This backstory, then, easily explains Zelda’s inventing evidence: the pre-existing jealousy affects the Bayesian updating by making the ‘affair’ hypothesis more probable than it is. This is a sensible idea, but it does not explain all cases of unwelcome self-deception. I do not see how one could sensibly fill in the details of Heartbroken Parents to make it consistent with the FTL. What kind of an event could make the hypothesis ‘We are not responsible for our child’s death’ a more costly error?

All variations of the Melean view raise similar concerns. Scott-Kakures (Reference Scott-Kakures2001, see Reference Scott-Kakures2000), for example, suggests that self-deception makes the agent’s goals and interests, such as avoiding danger, more likely to be fulfilled than not. The modification of the Melean view is in that the costly error is not measured with respect to falsely accepting/rejecting that p but rather with respect to the agent’s focal goals and interests, which may be promoted even by falsely believing/rejecting that p. A true belief may sometimes be a costly error, as in the case in which there is a danger of your partner having an affair and making you believe that the partner is having an affair will help you to prevent this from happening. On this view, that is, by acquiring the unwelcome belief, self-deceivers make sure that they will take all necessary steps to avoid the unwelcome state of affairs (see, Funkhouser, Reference Funkhouser2019: 235).

A similar hypothesis elegantly explains many cases of the self-deceptive imposter syndrome, a false, unjustified belief that one is successful in a highly competitive job due to luck. The explanation is that, because they believe that they lack the relevant ability but that hard work can compensate for this shortcoming, people with this syndrome work harder and thus perform better. Their negative appraisal of their own qualities provides additional motivation by signalling the need for more effort, and this additional motivation is both advantageous in the relevant context and outweighs the relevant affective disutility (Gadsby, Reference Gadsby2022).

Notwithstanding its merits, this view does raise some concerns if we try to use it as a general account of the origin of self-deception. Say that an athlete systematically deceives herself into believing that she is injured before important competitions – call this Hurt Athlete. Acquiring this belief systematically undermines her performance or even prevents her from competing rather than serving her best interests. Appealing to her perceived primary error requires explaining why the perceived error would be to believe that she is hurt. If the competition is important enough, many athletes compete even while hurt (often under severe painkillers), thereby risking serious injuries. In this case, the utility of the false belief not only does not outweigh the disutility of the true belief, but it only generates disutility.

We could say Zelda, John, and Heartbroken Parents involve a malfunctioned self-protective mechanism, an adaptive disposition that occasionally backfires. However, this position begs the question – it takes it as true that self-deception is a self-protective mechanism, which is exactly what is at stake here – and cannot explain Hurt Athlete since she harms herself systematically. Nevertheless, some cases seem to fit this hypothesis. Consider Barbara.

The unwelcome belief increases Barbara’s chances of preserving her house. If Barbara falsely believes that she has left the stove on, the cost is relatively low (she will return only to find that all is well) but, if Barbara falsely believes that she has not left the stove on, the cost is extremely high – her house might burn down (Scott-Kakures, Reference Scott-Kakures2001: 323). Therefore, this mechanism obviously serves her goals and interests, and there is no need to think that Barbara’s psychology makes her in any way more prone to self-deception than other people, which is a concomitant of Mele’s proposal, for instance.

However, why think that Barbara’s behaviour involves self-deception? After all, she cannot remember turning the stove off and she was absent-minded that day. Her belief, though false, is neither epistemically unjustified nor irrational, and the function of her behaviour is not to mislead. Levy (Reference Levy2016) gives us three possible interpretations of this behaviour. According to the first, Barbara’s thought would not be a belief but rather a default thought, a subpersonally generated simulation similar to imagination (Gerrans, Reference Gerrans2014). The default network is a mechanism of testing thoughts for consistency and adequacy, which in turn may qualify them for representational states such as beliefs. When unsupervised, default thoughts may occupy the role of representational states and make people act on them: people may think that they believe that they have left the stove on and go back to turn it off. Default thoughts may equally become beliefs if the mind judges that the thought is reasonable, which would be the second interpretation of the case. Accordingly, it could be that Barbara’s fear triggers the relevant default thought, ‘My house may burn down’, which, being judged as reasonable according to prior beliefs and available evidence, gets passed as correct and adopted as a belief. Finally, it could be that the fear is caused by Barbara’s representation of the stove as on; this fear generates anxiety, which then directly influences her behaviour. Each of these explanations is more plausible than the one that says that Barbara deceived herself into believing that she left the stove on.

Nelkin (Reference Nelkin2002) and Funkhouser (Reference Funkhouser2005) argue that both welcome and unwelcome self-deceivers want to believe the proposition they end up with. Welcome self-deceivers want this proposition to reflect the state of affairs, while unwelcome self-deceivers want their belief to aid in preventing this state of affairs from occurring. Unwelcome self-deceivers, thus, want to believe the unwelcome proposition or, at least, to have some first-person qualities associated with such a belief, but they do not want that what they believe is true. Being the jealous, insecure type, Zelda wants that Scott is not having an affair and, out of caution, wants to believe that Scott is having an affair (Funkhouser, Reference Funkhouser2005: 298) – since believing what she does not want to be true will make her more vigilant. However, postulating this kind of motivation is theoretically problematic, and the function of this motivation is unclear. Zelda and Barbara can perform all cautionary actions even without believing the unwelcome proposition (McKay and Dennett, Reference McKay and Dennett2009: 500–501). Therefore, it does not make sense for them to want to believe it and cause themselves unnecessary stress.

Perhaps the right kind of protective mechanism should be based on the wish to believe that one is at risk of getting cancer (losing one’s job, failing the exam) or, analogously, that one is at risk of losing her husband’s affection. This modification may explain some variations of Zelda and John, but it is still too narrow. Take, for example, Hurt Athlete. Believing that she is at risk of getting hurt right before the big competition will reduce her chances of success, which systematically defeats the purpose of the supposed protective mechanism. Why would this athlete want to believe that she is hurt? It makes more sense for her to cause herself to believe that she is not as good as she is, which might prompt her to train harder (Funkhouser, Reference Funkhouser2017: 237, n. 24; Gadsby, Reference Gadsby2022) or to lower her expectations (Funkhouser, Reference Funkhouser2019: 236). Finally, the Nelkin–Funkhouser hypothesis cannot explain the behaviour of the parents who blame themselves for their child’s death. The unwelcome belief will not help them to take all necessary steps to avoid the unwelcome state of affairs.

The only cases in which believing what one does not want to be the case but in which the belief is systematically serving the person’s interests are Barbara and some involving the imposter syndrome. However, because Barbara does not involve self-deception, all we are left with are some cases of the imposter syndrome, which suggests that furthering the self-deceiver’s interests in this way is more an exception to a systematic practice rather than a sign of a systematic practice.

The final explanation says that the self-deceiver’s emotional state primes relevant processing systems to gather evidence in a biased fashion, which – depending on the state – sometimes produces welcome and sometimes unwelcome beliefs. The welcome self-deceiver cannot face some unwelcome reality and so she turns her head away, whereas the unwelcome self-deceiver cannot stop thinking about some unwelcome reality and so she treats some data as evidence that what she fears of is true (similarly Lazar, Reference Lazar1999; Bach, Reference Bach, Beckermann, McLaughlin and Walter2009; Galeotti, Reference Galeotti2018: 52–55). In unwelcome believing, the person’s anxious emotion makes her incapable of getting her mind off the subject. She, therefore, dismisses information she can easily put out of her mind and concentrates on information she cannot put out of her mind, which makes the dreadful possibility more realistic than it really is. However, because the function of these mechanisms is not to mislead, this proposal cannot be used in an analysis of self-deception. Also, the priming mechanism does not apply to non-welcome self-deception (§4.3) since these self-deceivers are emotionally neutral, I am not sure how it could apply to Heartbroken Parents, and it sits uneasily with cases in which self-deceivers consistently probe the unwelcome hypothesis (Scott-Kakures, Reference Scott-Kakures2000: e.g., 361–362).

In conclusion, the idea that self-deception systematically brings people some benefit is rather problematic. However, before abandoning it, I will first try to refine it.

4.2.2 A Possible Refinement

All the solutions discussed thus far say that people deceive themselves because they somehow benefit from this. They only differ in what kind of benefit the person receives. So, perhaps we could simply say that people engage in self-deception because it benefits them in some sense, period. And here is one way in which this proposal can be cashed out as a systematic theory.

At least three factors determine why people engage in self-deception in some contexts but not in others and in a specific direction (welcome, unwelcome). The first is how the person perceives herself, namely, in a positive or negative light. The second factor is how she assesses the relevant context and what is at stake in it: low stakes mean low motivation, and low motivation means low probability of self-deception. And the third is how the utility of the relevant beliefs figures in relation to (1) and (2); namely, what is the perceived utility of the belief in the relevant context. I will address these each of factors in turn, but briefly – since they have already been widely discussed.Footnote ⁵

Some studies suggest that most people have unrealistically positive views of themselves and that they hold unrealistic optimism about the future (Taylor and Brown, Reference Taylor and Brown1988; Taylor et al., Reference Taylor, Skokan and Aspinwall1989; Trivers, Reference Trivers2011: 15–18). For instance, most people think that they are in more control of their life than they in fact are and they take more credit for success while denying responsibility for failure (Langer and Rodin, Reference Langer and Rodin1976; Snyder et al., Reference Snyder, Irving, Sigmon, Holleran, Montada, Filipp and Lerner1992; Kenny and Kashy, Reference Kenny and Kashy1994). This bias is unconscious (Epley and Whitchurch, Reference Epley and Whitchurch2008), but it is not always inflationary. People with negative self-views preferentially seek negative self-evaluations, and some of them even divorce spouses who perceive them in an overly positive manner (Swann, Reference Swann, Suls and Greenwald1983).

This discrepancy in self-evaluations explains why some people cause themselves to falsely believe that they are not losing hair and some that they are overweight or that they lack some important abilities, but it is insufficient for a general account: these self-evaluative biases are not hallmarks of certain types of personality. Therefore, even though some people may be more prone to self-deception than others, we should not think that there is something like a self-deception-prone type of personality. In different contexts and with different things at stake, the same person may exhibit different biases or exhibit no biases whatsoever. A man may absolutely dread losing his hair as an adolescent and merely find it unwelcome later in his life, since he no longer thinks that losing hair will affect his social status. Also, consider the results of a study conducted by John and Robins (Reference John and Robins1994). Their participants were MBA students at the University of California, Berkeley, who were simulating the meeting of a compensation committee in a large company. Some 35 per cent of the participants showed a self-enhancement bias (positively inflating one’s self-understanding), 15 per cent exhibited a self-diminishment bias, while the remaining 50 per cent were rather accurate. I think that the fact that half of the participants did not exhibit any bias makes sense on the thesis that performing better or worse than others had no motivational effect on them. A likely explanation is that they did not see this as a competition but rather as a task calling for teamwork. This study, then, suggests that the direction of the bias depends not only on one’s personality but also on what one takes to be at stake in that context. This is why the second factor is rather important.

The third factor explains how self-deceptive belief figures in the ‘why X deceived themself in this particular way in context C’ equation. For example, in some instances of the imposter syndrome, ‘the domain is one where the individual believes that effort can substitute ability, such that a belief that one lacks ability will have a motivating (rather than demotivating) effect’ (Gadsby, Reference Gadsby2022: 253). So, the utility of the relevant belief is in that it provides a beneficial additional motivational boost for the agent that will increase their chances of success. This explains the behaviour of those people who negatively see some of their abilities relative to some domains of their life in certain contexts but, in general, have positive self-views. Some people, however, have negative self-views and they preferentially seek negative self-evaluations. These people, then, with respect to the same domains of their life and the same contexts, may just need reasons to allow themselves to fail, a reason that will excuse their lack of trying, which counts as having utility for them. And this may happen even to people with positive self-views when the cost of additional effort may be a price too high to pay. If I need to invest significant resources to obtain X in C and the probability of obtaining X in C is insufficiently high considering the cost, deceiving myself into believing that the probability is substantially lower than it actually is would allow me to abstain from trying to get X in C and thereby preserve my resources for some future endeavour or maintain my overall wellbeing (which would be significantly affected by my failure).

This short exposition yields the following explanatory framework: (i) people engage in self-deception to secure their focal benefit (universal kind of result) and what determines whether this benefit will be secured by acquiring a welcome or unwelcome belief is (ii) how the person perceives herself and the relevant context (i.e., what is at stake), and (iii) which belief generates greater perceived utility – relative to (i) and (ii). Notice, one’s perspective on the issue – condition (ii) – is vital in this equation since it determines both whether the person will engage in self-deception or not and whether the product of self-deception will be welcome or unwelcome. Let us now apply this framework, call it refined self-interest view, to the cases discussed.

Zelda and Scott’s marriage is going through a difficult phase: the ‘affair’ hypothesis explains Zelda’s marital problems while shifting blame to Scott, thereby protecting Zelda’s positive self-image; this is the utility of the self-deceptive belief. But it could also be that Zelda is avoiding the thought that the marriage is going through a difficult phase because Scott does not find her sexually attractive. As a result, ‘Cheating fits Scott’ means that his faulty character rather than her lack of attractiveness caused their marital problems. In both variations, what explains why she engages in self-deception (condition 1 of my functional analysis of self-deception) is her need to protect her positive self-image (focal benefit) in this context and the fact that this need outweighs the cost of suffering some harm. Her behaviour is not so strange after all.

Hurt Athlete is best explained along the lines of Quattrone and Tversky’s (Reference Quattrone and Tversky1984) analysis of self-handicapping strategies. People engage in self-handicapping strategies in order to ascribe the failure to their handicap and not to their lack of ability to succeed: ‘by drinking or taking drugs [before the upcoming IQ test], any level of intellectual performance would not destroy the belief that one is basically bright, for even failure could be attributed to the debilitating effects of alcohol’ (Quattrone and Tversky, Reference Quattrone and Tversky1984: 248). The need to avoid explaining the failure to one’s lack of ability to succeed, that is, explains why one engages in self-deception just as the need to deter predators explains why the snake plays dead. The athlete’s belief that she is hurt protects her positive self-image by giving her a reason not to compete or an excuse if she fails to meet her expectations at the competition, which explains why she systematically deceives herself.

Funkhouser and Hallam (Reference Funkhouser and Hallam2024: §3.3) argue that self-handicapping strategies require elaborate unconscious knowledge and an unconscious ‘plan’ to create an excuse for an anticipated failure or suboptimal performance. On their view, then, these cases of self-deception require a deeply divided mind and a kind of intention to mislead oneself. Allow me to briefly suggest a simpler solution that does not require positing subconscious deceptive intentions. This behaviour could predominantly be an automatic, pre-conscious strategic response to a specific stimulus analogous to a firefly’s response to a specific signal (condition 1 of my analysis). Relative to the three factors I discussed, some athletes systematically encourage themselves before the competitions, whereas some systematically come up with excuses. Rather than unconsciously evaluating the evidence, context, the probability of failure, and the plausibility of the excuse, these self-deceivers, just like their non-self-deceiving peers, may simply react strategically: in context C, they perform behaviour φ where, for some athletes, φ-ing is encouraging oneself, and for some others, it is coming up with excuses. This is another example of how the fact that it can easily explain how people strategically and without a plan deceive themselves is an important advantage of the functional view.

The case of the man who thinks that the government is spying on him can be explained along the lines of Bentall’s (Reference Bentall, David and Cutting1994: 353) thesis that persecutory delusions protect the individual against chronic feelings of low self-esteem. This man thinks that he is important enough for the government to see him as a threat. The paranoia gives him a sense of value and feeds his self-regard (condition 1), thereby securing his focal benefit, which explains why he deceives himself.

The proposed account seems rather successful: while we have identified one general reason to engage in self-deception, we can apply a case-by-case analysis of the phenomenon by reasonably filling in some details. However, this is still not good enough: while this hypothesis explains many or most cases of self-deception, it cannot be a general analysis of self-deception. Some cases of self-deception are non-welcome (so, the belief’s utility plays no role in its aetiology), and a non-negligible class of self-deceptive behaviour systematically does not benefit the agents in any sense.