• 1 Raz 1999 [1975]: 36.
• 2 Raz 1999 [1975]: 39.

1When faced with a certain action, human beings often have reasons for and against that action. Imagine a person is hidden in the bushes with a rifle as a threat approaches. Since her life matters, she has a reason to remain hidden or to flee. But because her people’s safety matters too, she has a reason to fire when the threat comes up. This is an example of first-order reasons for action, i.e., reasons to do or not to do a certain action (resting). When a conflict between first-order reasons occurs, the conflict is resolved according to the relative strength of the conflicting reasons.1 According to Raz, alongside first-order reasons we have second-order reasons, which are any reasons to act on certain other reasons or to refrain from acting on certain other reasons. When a second-order reason is a reason to refrain from acting on certain other reasons, that reason is an exclusionary reason, where refraining from acting on the excluded reasons is compatible both with not doing the act and doing it but not for the excluded reasons.2 Regardless of whether she has reasons to shoot or not, the woman with the rifle may have reasons for not considering the importance of her life in making her decision.

• 3 Raz 1999 [1975]: 191; for the expression “pre-emptive reasons” see Raz 1985: 42, 46; cfr. Gur 2018: (...)

2In Raz’s thinking, the notion of exclusionary reason is crucial for understanding the concept of legitimate authority and then of legitimate law. According to this approach, in fact, authoritative directives are pre-emptive reasons, which means that they are both a first-order reason for a certain action and a second-order reason to exclude some other reasons –— which may count against that action (or possibly in favour of it).3 This means that when an authoritative directive prescribes the woman to shoot, the directive is a reason to shoot, but this reason is not just another reason to be added to the other reasons and to be balanced with them. Rather, the directive excludes (displaces, defeats) some of the other reasons, like the relevance of her life.

• 4 Gur 2018: 16.
• 5 Raz 1986: 53.
• 6 Raz 1986: 61-62.
• 7 On the assumption that the actual location of the threat is one of the reasons inside the scope of (...)
• 8 Raz 1999 [1975]: 38.
• 9 This way of referring to Raz’s theory of authority has been introduced by Vassiliou 2022: 843.

3We may call the first-order reasons existing before the directive has been issued “underlying” or “background reasons”.4 According to Raz’s service conception of authority, whether an authority is legitimate and produces pre-emptive reasons depends on whether people will conform better to the balance of underlying reasons if they try to follow the directive issued by the person or body claiming authority, rather than if they try to follow the balance of underlying reasons directly.5 The function of authorities is to make people do the right thing. Still, the directive issued by a legitimate authority qua exclusionary reasons must be obeyed even when wrong.6 For tactical reasons it may be better that the soldier does not shoot at the enemy at that moment. However, the directive supersedes the underlying reasons and the soldier ought to shoot.7 Since authoritative directives replace some of the background reasons, we may say that they are reasons not to act on the merits of the case at hand — reasons “for doing what you were ordered regardless the balance of reasons”.8 The idea that authoritative directives are pre-emptive reasons is the content of what we may call “the exclusionary model”.9

• 10 Gur 2018: 15 and the literature cited; see also Hurd 1991: 1641 ff.; and Schauer’s rule-sensitive p (...)

4The exclusionary model is then contrasted with the weighing model. The weighing model does not entirely deprive authoritative directives of normative relevance. The fact that there is an authoritative directive in favour of doing x provides, according to this model, a first-order reason in favour of doing x.10 The first-order reason may be a very strong reason and may tip the balance, but, unlike in Raz's exclusionary model, it will be a reason that must be balanced with reasons of the opposite sign.

• 11 This idea in the Italian context has been illustrated by Bobbio 1984: 157 ff.; and more recently by (...)

5One of the rationales of the exclusionary model involves a distrust about human decision-makers.11 To the extent that well-motivated people are prone to error in weighing reasons for action, it is better that they just try to follow authoritative directives. This distrust in the judgement of moral agents is an argument in favour of the exclusionary model over the weighing model, since in the latter model the agent is still asked to weigh the underlying reasons and authoritative directives as first-order reasons.

• 12 I am not suggesting that the struggle against biases was central in Raz’s theory. Raz argued that t (...)

6One of the reasons for distrust in the practical performance of agents lies in their susceptibility to bias. The law and authoritative directives in general can be seen as a tool to overcome psychologically distorting factors.12 As Gur points out,

• 13 Gur has also analysed the problem of biases in relation to the exclusionary reason model (see Gur 2 (...)
• 14 According to this analysis the exclusionary model is a normative model about the type of reasons we (...)

8Against this background, the exclusionary model promises to better shelter agents from their biases.13 But the extent to which the exclusionary model actually succeeds in achieving this goal is unclear, and the purpose of this paper is to address this issue.14 Sections 2 and 3 are dedicated to the premises of the argumentation. In particular, Section 2 delimits the target, clarifying which interpretation of the exclusionary model is under consideration, while Section 3 delimits the type of biases and experiments challenging the exclusionary model so understood. The following sections constitute the attacks on the exclusionary model, based on experimental evidence. They focus respectively on our inability to intentionally disregard certain factors, in general (Section 4), in the moral sphere (Section 5), and finally in the legal sphere (Section 6). The attack combines a general framework on our inability to intentionally ignore relevant or quasi-relevant information — studies on “mental contamination”; Jonathan Haidt's research on the conflict between moral reasoning and moral intuitions; and studies on motivated reasoning. While the main aim is to support the critique of the exclusionary model, this juxtaposition of empirical research may be useful for other purposes as well. In particular, it could weaken the premises of Haidt’s theory about ex post rationalization strengthening its conclusions, and lay the groundwork for further analyses about the relations between law and morality. The article concludes with a discussion of the conclusions reached so far and the prospects for new research that they open up.

• 15 See, also, Raz 2009a [1978]: 25-27. The position according to which obedience doesn’t imply surrend (...)

10So Raz distinguishes the sphere of action from the sphere of deliberation.15 The conceptual premise is that acting against her judgment is not surrendering our judgment, and the main idea is that, to obey authoritative directives, we must act on the pre-emptive reasons stemming from the directives, while we are not required to “stop thinking” about the issue. After authoritative directives have been issued, it is no longer up to us to decide how to act on the basis of a balance of reasons, but we are allowed to form our own judgement, as long as this judgement remains unrelated to our behaviour.

• 16 I borrow this expression from Rachlinski & Wistrich 2017: 94.
• 17 A clarification might be helpful to avoid some misunderstandings. According to Raz's preferred appr (...)

14In other words, the exclusionary model — according to this interpretation, which will be problematized only at the end of this paper — requires us to exclude reasons for actions by performing a mental task that we will call “the exclusion task”, where by “exclusion task” I mean a conscious and intentional mental action by which the agent neutralises the relevance of certain mental states. The question is whether this model of reasoning is safe, i.e., whether people can compartmentalize their thoughts.16 Indeed, it would be an empirical challenge to the exclusionary model if people are unable to compartmentalize their thoughts. If the inability to compartmentalize our thoughts is the inability to perform the exclusion task, this means that in many cases we would not be able to live up to the demands of the exclusionary model.17

• 18 Danzinger, Levav & Avnaim-Pesso 2011.
• 19 In general, see Kahneman & Tversky 1974.
• 20 Wistrich & Rachlinski 2017: 94.

17In starting the analysis of the psychological challenge to the exclusionary model, we need to understand which types of biases are more pertinent and which are less pertinent for this challenge. Irrelevant factors can determine our judgment by acting completely behind our backs. A first example is Danziger, Levav, and Avnaim-Pesso’s famous work on the influence of food breaks on parole judges. They found that the 65% approval rate of parole applications after a meal declines steadily in the next two hours, and drops to about 0% before the next break.18 The anchoring effect provides another useful example. Anchoring describes a phenomenon according to which numbers inserted in a scenario are taken as relevant for the solution of a numeric task, even though they are totally irrelevant.19 In an unsettling study, for example, judges have been asked to quantify a fine for the violation of a noise-ordinance by a nightclub. Some judges were familiar with a version of the case in which the nightclub was identified as Club 11866 (based on the street address), while others read a version of the story in which the club was identified as Club 58. The number of the identification anchored the judges' response, and those who identified the club with the higher number were more likely to impose a higher fine.20

• 21 I.e. our ability to exclude what we take to be a reason (see footnote 17).

18Experiments of this kind do not specifically challenge the exclusionary model. For an experiment to challenge the exclusionary model, it must show that people fail to exclude reasons, that is, they fail to successfully complete an exclusion task. But in cases like these, because the disturbing factor is totally irrelevant, and the agent may well be unaware of its dangerous influence, we cannot say that the agent recognizes to be dealing with an exclusion task. Since these cases challenge our faculty to follow both models, they are unable to challenge any one model in particular. The crucial question then is what psychological distortion, and thus what experiments, can challenge our ability to exclude reasons.21 First, to challenge our ability to perform the exclusion task, we must have experiments focusing on biases that, although identified by the agent, are not corrected; that is, experiments in which the agent tends to be aware of the biasing nature of the situation in which she is operating. So, the first feature of the psychological distortions we are interested in is the awareness element.

• 22 Again, for “information that is reason” I mean “information that the agent takes to be a reason”, s (...)

19Second, strictly speaking, our ability to perform the exclusion task is challenged if the design of the experiment shows that the biasing information that must be excluded is considered relevant by the agent, i.e., if the information is considered (from her perspective) to be a reason — relevance element. If we showed that moral agents can exclude information over facts that counts as reasons,22 but not information over facts that by no means constitute reasons, we would not have challenged the exclusionary model. We would only know that both the weighing model and the exclusionary model often allow completely irrelevant factors to influence our judgement. But the relevance of the exclusionary model is weakened when it is shown that moral agents are unable to exclude information over facts that count as reasons, since the weighing model, on the contrary, does not stipulate that this information should be excluded.

• 23 Therefore, the idea of quasi-reasons should not be confused with prima facie reasons.
• 24 The above applies in relation to the agent's point of view, which is what is of interest here. From (...)

23Before moving on, one last consideration. So far, I have divided factors, information, and circumstances between relevant and irrelevant — reasons and non-reasons. However, plausibly there are irrelevant factors, information, and circumstances — therefore non-reasons, after all — that somewhat resemble relevant ones: quasi-reasons. By “quasi-reasons” I mean factors that have the structure of a reason, in a sense in which random anchoring numbers do not.23 The fact that my boss was aggressive may be a factor that causes me to challenge him. Postulating that from my point of view, my boss's aggressiveness is not a reason to challenge her, the fact remains that I view this fact as somehow relevant for practical purposes. Phenomenologically speaking, even though for me my boss’s aggressiveness is not a reason to challenge her, I may realistically conceive of a person for whom this fact is indeed a reason; while it is extremely hard to imagine someone attaching practical values to random numbers.24 A bad reason, in this sense, can be a quasi-reason, whereas random numbers never can be. Why distinguish between two types of irrelevant factors, introducing the concept of quasi-reasons? Well, to evaluate the performance of our ability to exclude reasons, one must also consider the case of quasi-reasons as a challenge to the exclusionary model. Just as the burden of proof of the element of awareness can be lightened in the presence of relevant information, so can it be, though perhaps to a lesser extent, in the presence of information pointing to quasi-reasons. Even if for me sexual orientation is not a reason to condemn someone, it is still not the same as random numbers because for me sexual orientation is a salient factor, and I am more likely to be aware that I am faced with an exclusion task.

• 25 See the title of Part B – “Anchoring, Contamination, and Compatibility” – of Gilovich, Griffin & Ka (...)

26In and of itself, this is not an interesting definition, given its level of generality. Moreover, although this expression was previously adopted by some of the most prominent experimental psychologists,25 it does not enjoy great popularity today. Nevertheless, the idea of contamination serves to emphasize the agent’s struggle to debias her thoughts:

• 26 Wilson & Brekke 1994: 128-130, 131-133. I will not try to establish here whether the problem for th (...)

28What is relevant for our purpose is that according to this account the detection of the bias and the motivation to correct it are not enough, since for mental contamination to occur it is sufficient to fail in each of the subsequent steps. In particular, the agent may confuse the source of her judgment, believing that she got rid of the bias and that the source of her judgment are undefeated reasons only, whereas the truth is that the bias is still the real source of her judgment — source confusion26. It is easy to see how the exclusionary model may fall prey to the source confusion. As we have seen, the exclusionary model seems to be based on our ability to compartmentalize our thoughts. The ability Raz seems to be relying on when he says that the obedient agent is not required to surrender her judgment, but only to act on the basis of the authoritative directive. Source confusion is the negation of this ability.

• 27 See Gilbert 2002: 180; Wistrich, Guthrie & Rachlinski 2005: 1264.

29The failure of excluding information that is or was somehow relevant for our deliberation is well analysed in studies about belief perseverance.27 According to a popular and intuitive view, which may be traced back to Descartes, understanding and believing are separated mental processes. When a rational agent understands a certain piece of information, it does not mean that she accepts that information as true; on the contrary, she will believe the information only after weighing the arguments for and against its credibility. The Cartesian model of belief formation is contrasted with an opposite model, inspired by Spinoza, according to which understanding a piece of information and believing it are a single mental operation: when a human being understands a piece of information, she begins to take it as true — that is, to believe it — and then decides whether or not to “unbelieve” it. But unbelieving what we believed before is hard because, according to belief perseverance, once the agent has been exposed to a piece of information she integrates the information in her web of beliefs, fabricating explanations revolving around the belief, to the point that if the information is discredited or becomes useless for other reasons, the explanations that the agent has fabricated in the meantime persist, making the belief persist. One case of mental contamination is provided then by the difficulty of un-believing what was previously believed.

• 28 Damaška 2003.

30Before turning to consider two experiments on belief perseverance it is useful to introduce two notions. Among the reasons to disregard information, it is possible to distinguish between epistemic and extra-epistemic reasons. Epistemic reasons have to do with fact-finding. So, when we have epistemic reasons for ignoring certain information, it means that we have reasons for ignoring information because it is wrong — it does not lead to truth. But while truth certainly has value, it is not the only value, and we may have reasons to ignore information even if it is epistemically valid: extra-epistemic reasons.28

• 29 Golding & Hauselt 1994.

31The first experiment concerns the evaluation of a teacher. 29Participants were asked to rate a teacher assistant on a scale from one (least nice) to ten (nice). In the first experimental group, some experimenters in disguise gave participants negative information about the teacher. In the second experimental group they gave negative information too, but afterwards they asked participants to disregard the information for extra-epistemic reasons, such as “I probably shouldn’t have told you those things.” In the third experimental group they gave negative information, but afterwards they asked participants to disregard the information for epistemic reasons, saying that they were confused and they were referring to another person. Finally, in the control group the participants did not hear any negative information about the teaching assistant. Two things are interesting about this study. First, the evaluations by the first and second group were very similar, and both were very different from the control group. This shows that suppression was totally ineffective in the second group. Second, both the second and third groups had been exposed to the information and were asked to suppress it. However, the rating provided by the third group was statistically higher than the one by the second group. This means that the reasons supporting suppression are relevant to its effectiveness: in this case, epistemic reasons worked better than extra-epistemic reasons.

• 30 See generally Ross, Lepper & Hubbard 1975.

32But we should not place too much hope in the reasons given in favour of suppressing information. In a second relevant experiment, subjects were instructed to read some suicide letters. They were informed that some were written by the experimenters and asked to identify the genuine suicide letters. Their performances were then ranked by the experimenter. Some of them performed well, while others poorly. Later, in a second stage, the experimenter revealed that all the letters were fake. In a third stage the experimenter asked the participants to rank their ability to recognize real suicide letters. The results of this experiment showed that those who were evaluated positively at the beginning of the experiment, expressed that they would have performed well even in a real task, while those who were given negative evaluations at the beginning answered that they would have performed poorly.30 So, even if they had epistemic reasons to ignore the evaluation, the information they acquired persisted in their thoughts.

• 31 It is possible that in Raz's theory it would be more appropriate to say that reason has been cancel (...)

33The idea of belief perseverance challenges the exclusionary model insofar as it casts doubt on our ability to exclude reasons. From the agent’s perspective, a certain fact looked like a reason for adopting a certain judgment, but even after the agent discovered that the fact in question is not a reason to adopt that judgment, the fact continues to influence their judgment.31 The fact is no longer a reason, but its ghost continues to influence the agent — a phantom reason.

• 32 The expression “layered morality” is not taken from Haidt's theory, but I am convinced that it is u (...)

36A clarification needs to be made in this regard. It might seem unusual to consider people's morality as a distorting factor. Certainly, a person may obey moral principles that we consider to be wrong, but this is not the sense in which morality can be said to be a distorting factor. We must remember the definition proposed above: for there to be mental contamination we must have “an unwanted judgment (…). By unwanted, we mean that the person making the judgment would prefer not to be influenced in the way he or she was.” In other words, the agent's morality must be an undesirable influence not from the point of view of the external observer, but of the agent himself. We commonly conceive biases as obstacles for the right thing to do, that is for the fulfilment of our moral requirements. How, under this assumption, could a person's morality be the source of a judgment undesirable to her? It can be if we imagine the agent's morality as a layered morality, following, for example, Jonathan Haidt's suggestions.32

• 33 Haidt & Bjorklund 2008: 203. Haidt will distinguish later the fairness foundation and the liberty/o (...)

37In a nutshell, Haidt holds that people’s intuitive morality is articulated in six moral foundations.33

• 34 Haidt 2012: ch. VII.

1) The care/harm foundation
2) The fairness foundation (fairness as proportionality)
3) The liberty/oppression foundation
4) The loyalty/betrayal foundation
5) The authority/subversion foundation
6) Sanctity/degradation foundation34

• 35 See Haidt 2001, and Haidt 2012: ch. II.

38Along the lines of dual process theories,35 Haidt believes that our moral practice should be analysed through the distinction between intuition and reasoning. Moral judgment is the result of our intuitions, while reasoning comes after the judgment has been adopted. People are generally led to think that their judgment is the inferential product of reasoning because, while the intuitive part of moral judgment is unconscious, reasoning is a conscious activity, and it is the only part of the process which is experienced by the agent.

• 36 This aspect is discussed at length in Rocché & Ubertone 2024: par. 2.
• 37 See in particular Haidt & Hersh 2001.
• 38 “Why Good People Are Divided by Politics and Religion” (italics added): This is the telling sub-tit (...)

39According to this framework, since our moral judgments are very often causally determined by our intuitions, and moral foundations are the clusters of our intuitions, moral foundations dominate our moral responses. In some cases, our moral intuitions may be seen as a sort of heuristic, anticipating the judgments that we would adopt through the effortful process of moral reasoning. But Haidt is generally interested to another type of situation, in which there is a discrepancy between the moral foundation determining the judgment and the moral principles featuring in our justification.36 For example, some people, in evaluating a case of incest in which siblings took different contraceptive measures, are driven by the sanctity/degradation foundation to condemn this harmless but offensive conduct. When they are asked to provide reasons for their condemnation, they resort to shaky justifications related to the harm principle, such as the risk of having offspring with some malformation due to endogamy, even though the story clearly stated that the siblings took measure to avoid this risk. Reasoning is a form of ex post rationalization. One ingredient of this mismatch between the factor causing the judgment and its justifications is the pressure of social context to conceal some of her moral foundations.37 But the explanatory centrality of social pressure does not mean that the agent is insincere when she refers to the harm principle to support her judgment: to the contrary, because the real source of the judgment remains unconscious, the agent honestly believes that her condemnation of incest is based on her commitment to the harm principle.38

• 39 Kunda 1990: 480, 482 ff. According to Kunda’s framework the phenomenon we are considering is the “d (...)
• 40 Ditto, Pizarro &Tannenbaum 2009.
• 41 Kunda 1990: 483; see for the introduction of the expression Pyszczynski & Greenberg 1987.
• 42 Sood & Darley 2012: 1322; but the closeness between the two models is emphasized by Ditto, Pizarro (...)

40The idea of morality as a distorting factor in practical judgment is also at the heart of a tradition of empirical studies that will be taken up in the next section: studies on motivated cognition or motivated reasoning. In motivated reasoning the agent’s preferences distort the process of belief formation.39 Within this general scheme, the agent’s morality is one among the relevant preferences capable of orienting her reasoning, leading to a motivated moral reasoning.40 As in the case of Haidt's intuitionism, in the case of motivated reasoning, the agent operates in good faith under “the illusion of objectivity”, to use the expression employed by psychologists.41 One of the differences with Haidt’s account that is sometimes emphasized in studies about motivated reasoning, relates to the limits of the agent's ability to form credible justifications for the conclusions she wishes to reach.42

• 43 “Moral emotions are one type of moral intuition, but most moral intuition are more subtle; they don (...)

44To challenge the exclusionary model, we can dispense of the relevance element, as long as we have the awareness element. But here there is another problem: Haidt's theory of moral judgment is grounded precisely on the rejection of the awareness element. Indeed, Haidt hypothesizes that moral judgment is the result of intuition, which is equivalent to saying that the agent's moral foundations act unconsciously. It is significant in this regard that Haidt points to intuition and not emotion as the causal factor of judgment precisely because he wants to emphasize how often the mechanism that causes moral judgment is not perceived by the agent.43 Our morality would catch us off guard, and there would be no moment when we conceive of our task as an exclusion task.

• 44 Our personal assessment of background reasons: solitary moral reflection; or moral reflection in a (...)
• 45 See for example Schauer 2004; and Maroney 2011.

49Be that as it may, even if studies on practical judgment like Haidt’s were unable to challenge the exclusionary model as long as we remain within pure moral reflection,44 because we believe that in this context the agent is not sufficiently aware that she is faced with an exclusionary task, or because the excluded considerations are not reasons, a more convincing attack on the exclusionary model can be made once we move to the legal realm. In fact, the law and legal culture often explicitly require the agent, when applying the law, to exclude certain considerations that might seem relevant from a practical point of view.45

• 46 Fuller 1964: 81; Celano 2023: 214.
• 47 Raz 2009b [1977]: 213.

50As was mentioned in the opening section, the notion of exclusionary reasons is central for understanding the authority of law. A crucial aspect of the ideal of Rule of Law is that, in Fuller’s words, there must be a sort of “congruence between official action and the law”, or, following Bruno Celano’s analysis, that the law in some sense predetermines the decision to be taken in a particular case.46 For Raz “what the doctrine [of the rule of law] requires is the subjection of particular laws to general, open, and stable ones”.47 The basic problem — inherited from the previous discussion — is whether the agent is capable of neutralising her own morality in order to be able to correctly apply the authoritative directive, or whether morality contaminates the application of the law. The scheme of the moral contamination of law, which for simplicity's sake I will call “legal contamination”, can be summarised as follows:

• 48 See Moore 1989: 839 ff. As Moore notes, this continuity is one of the main differences between Raz' (...)

51The scheme seems to contrast moral reasons and legal reasons, i.e., the reasons for following the authoritative rule. This approach would contradict the spirit of Raz's theory of authority. For Raz there is a continuity between jurisprudence and moral theory, and the reasons to follow the authoritative directive (instead of acting on the balance of background reasons) are moral reasons too:48 the conflict between morality and authority must therefore be represented as a conflict between morality and morality.

• 49 Respect for authorities is also one of Haidt’s six moral foundations, and in this sense it is a com (...)

53We can distinguish three levels of morality. First, there is the unconscious morality that is at odds with the moral principles consciously held by the subject. In Haidt’s analysis, the moral foundation of sanctity and degradation would play this role in many peoples’ minds. Second, there is the conscious morality that concerns the background reasons. We are in this sphere when we try to do the right thing by balancing the various background reasons. Finally, there is our conscious morality, which recognises the existence of authority and exclusionary reasons. The transition from the second to the third morality is marked by the advent of epistemological and coordination problems, it is fuelled by the dangers of a homemade decision-making process of weighing reasons.49 Once authorities are recognized, the second morality is, from the point of view of the obedient agent, phenomenologically a kind of “shadow cabinet”: although we know that rules are fundamental and must be obeyed, we often simulate what we would have more reason to do in a world in which there were no legitimate authorities. According to Raz, this simulation — the installation of the shadow cabine is not precluded by the presence of legitimate authority, and this is why authorities do not require us to surrender our judgement. Having said this, in what follows, for expository convenience, the contrast between the first two moralities and the third morality will often be rendered as the contrast between morality (conscious or unconscious) and the law.

• 50 Wistrich, Guthrie & Rachlinski 2005: 1298 ff.; see also Wistrich & Rachlinski 2017: 95.
• 51 Information about the sexual conduct is in my terminology an example of “quasi-reasons”. In this ca (...)

58In the first experiment, Wistrich, Guthrie, and Rachlinski wanted to test the efficacy of a “rape-shield” statute, in the deliberative process of real judges. The rape-shield statute is an exclusionary rule that limits the admissibility of information on the chastity and sexual history of the persons involved in the case.50 The purpose of this provision mixes both extra-epistemic and epistemic concerns. On the one hand, the law wanted to avoid hostile examinations of the alleged victim by the defendant's lawyer. On the other hand, there are epistemic concerns about the admissibility of evidence concerning sexual history. Details of sexual history may be totally irrelevant, or they may be relevant, but in a very uncertain sense. According to some, for instance, promiscuous sexual conduct may suggest that the sexual intercourse was consensual; according to others, promiscuous sexual conduct is a favourable condition for the credibility of a rape report: sexually experienced persons may be more credible, since, in their long history of sexual intercourse, they have not falsely accused their partners of rape.51

• 52 Wistrich & Rachlinski 2017: 95.

59The story depicted in the plot of the experiment is that of a college student who is accused of having sexually assaulted a fellow student during a fraternity party. The complainant had been seen speaking with the accused male at the party before they were seen moving into a room together: apparently the girl was drunk and the boy was helping her walk. The complainant admitted to having had sexual intercourse with the defendant but denied having given consent. Evidence indicates that she immediately contacted the police after the event and that bruises consistent with a sexual assault were found on her body.52 Judges are asked whether they would convict the accused for sexual assault. But there is a crucial additional detail to take into consideration. In the suppression group, the experimenters added another piece of information. The defendant’s attorney tried to introduce the following evidence:

• 53 Wistrich, Guthrie & Rachlinski 2005: 1301.

60It was also added that “the prosecution has moved to exclude such evidence on the ground that it violates Arizona’s ‘Rape Shield’ statute . . . which forbids the introduction of evidence concerning a victim’s ‘chastity’ or ‘reputation for chastity’ in cases involving sexual assault”.53 The judges were then asked whether the information should have been excluded — as requested by the prosecutor — or admitted.

• 54 Wistrich & Rachlinski 2017: 95.

61We have three groups of judges to compare. Those in the control condition who never heard the testimony and knew nothing about the sexual history of the alleged victim; those who heard it and found it inadmissible; and those who heard the testimony and admitted it. The interesting result is that while the conviction rate was almost 50% among judges who were not exposed to the testimony, this rate was only 20% among those who heard the testimony but suppressed it. Statistically speaking, there was no difference with judges who read the testimony and admitted it: “In effect, it made no difference whether the judges who read the inadmissible evidence excluded or admitted it; regardless of their rulings, they relied on it”.54

• 55 As we said in the first section, the weighing model does not entirely deprive authoritative directi (...)

62At first sight the experiment seems to be a promising way of scaling down the advantage of the exclusion model over the weighing model. The way the experiment is designed seems to give an evocative representation of how subjects would behave if, on one hand, they acted on the basis of the balancing of background reasons or on the basis of the weighing model,55 or, on the other hand, if they acted on the basis of the exclusionary model.

• 56 The rule on inadmissibility of evidence has a special feature in that the operation prescribed by t (...)

65It seems that for many of the judges who excluded the evidence and convicted the defendant, the evidence unconsciously influenced their judgement. This would be a case where professionals – trained to apply the law – recognise that the law requires them to exclude certain information, are motivated to do so and believe they are fulfilling their duty, while they are disapplying the law: unintentional disobedience. If people are unable to perform the exclusion task when aware of it and motivated to do so, the advantage of the exclusionary model over the weighing model is diminished. This possibility deserves attention.56

• 57 As an anonymous reviewer noted, caution is called for when referring to empirical data, since even (...)

68What is missing is clear proof of the motivation to follow the exclusionary model rather than the weighing model or the simple balancing of background reasons. The fact that judges recognized that they are faced with an exclusion task and that they declared to have fulfilled the task, while we know they didn’t, is consistent both with the thesis that they have been unconsciously conditioned by the information, and with the thesis that they intentionally opted for balancing the reasons. In one case we are in front of a failure to follow the exclusionary model, in the other, it is the case that judges opted for weighing (what they believed to be) the background reasons, or for the weighing model properly called (background reasons + authority of law as first order reasons). This possibility is open, but taking it seriously is tantamount to saying that the experiment is completely unable to demonstrate anything particularly psychologically relevant.57 If we consider the experiment relevant — until someone proves the contrary — then we must consider it as a proof of a “failure to deliberately disregard relevant information”, that is a failure of the exclusionary model.

• 58 Sood & Darley 2012: 1321.

70According to the harm principle applied to criminal law, only harmful conduct deserves to be sanctioned through criminal punishment. The hypothesis studied by the two authors of the experiment in question, Avani Metha Sood and John Darley, is that when people want to punish a certain behaviour that is not harmful and they are presented with a legal constraint reflecting the harm principle, they tend to perceive the behaviour as harmful, expanding their conception of harm — the plasticity of harm hypothesis. Again, the manipulation is not conscious. Rather, people genuinely believe that their judgements on concrete cases are an application of the harm principle, when instead their perception of harm is inadvertently influenced by their goals — the illusion of objectivity.58 The experiment assumes, in other words, the occurrence of the previously mentioned phenomenon of motivated reasoning.

• 59 Sood & Darley 2012: 1325 ff.

71In the first phase of the experiment, a behaviour is identified that many people would like to criminalise even though they believe that this behaviour is harmless.59 One of the conducts selected by the authors in this first phase of the experiment is “going naked to the supermarket”. Going to the supermarket naked is thus a conduct that leads people to infringe the harm principle because according to the harm principle only harmful conduct may be criminalized.

• 60 Sood & Darley 2012: 1328 ff.
• 61 Sood & Darley 2012: 1328.
• 62 Sood & Darley 2012: 1328.

72The second part of the experiment is aimed at stimulating a process of motivated reasoning and proving the “plasticity of harm hypothesis”.60 People are divided into two groups, the experimental group and the control group. Both groups are presented with the story of the person going to the supermarket naked, but the experimental group is offered additional legally relevant information. They are told that “U.S. courts have decided that the government can impose a criminal penalty only upon conduct that is shown to cause harm”.61 In essence, it is stated that the courts have ruled that criminal law must always respect the harm principle: “the necessity-of-harm constraint”. This is not necessarily realistic, but the authors of the study make sure that the people tested believe that this is a true principle established by U.S. courts. In addition, both groups are given a definition of “harm”, whereby “Harm, for these purposes, is defined as injury to a person or persons that can be clearly demonstrated. There could be types of conduct that are wrong, but do not cause harm”.62 Both groups are then asked the following questions: (i) whether the government should repress the conduct under consideration through criminal law; (ii) whether the conduct caused a demonstrable harm; (iii) how harmful the conduct was.

• 63 Sood & Darley 2012: 1333.
• 64 Sood &Darley 2012: 1324 f.

73Well, if law had a causal impact on the decision, then (i) the rate of perceived harmfulness of the conduct should remain unchanged between the control and experimental group; and consequently (ii) the rate of criminalization of the conduct in the control group should plummet in the experimental group. Instead, the result of the experiment is that in the experimental group, the rate of criminalization stays constant even though people are subject to the necessity-of-harm constraint, and the number of harm reports more than doubled in the constraint condition as compared to the control condition.63 By being subjected to an argumentative constraint, people perceive conduct as harmful and criminalize it on the basis of this perception. This should prove the hypothesis according to which people, without being aware of it, are able to manipulate their conception of harm to achieve the desired result.64 We can say that law has a causal efficacy on people, but this efficacy does not concern the decision, which remains unchanged, as much as the reasoning that leads to the decision; reasoning that does not reflect the true reasons leading to the conviction and is therefore a form of ex post rationalisation.

• 65 See Sood & Darley 2012: 1342 ff.

74In this case, according to the Razian framework, the ruling of U.S. courts is a first-order reason not to criminalise harmless conduct and a second-order reason excluding reasons for and against criminalization. The structure of the experiment ensures that people are aware of the exclusion task they are facing. People's attention is drawn to the exclusionary task twice: first when they are told about the ruling of U.S. courts and second when they are given a conceptual scheme according to which there are wrongs that are not harms. Moreover, in this case it is quite plausible that the element of relevance is also present. It is plausible that not only does going to the supermarket naked trigger one or many of the subjects’ moral foundations, but also their conscious moral values, as was demonstrated through the first phase of the experiment in which the authors found conduct that for many deserves criminal punishment even if it is not harmful. Finally, it must be remarked that in this experiment the authors also attempted to prove the element of motivation, by proving that the subject responded “under the illusion of objectivity”.65

• 66 “Why not?” someone might ask. Given the indeterminacy of legal concepts, a reasonable person may in (...)

75On these grounds, the experiment questions the advantages of the exclusionary model over the weighing model. Thanks to the responses recorded in the first phase of the experiment, we know that for many people the result of the balancing of the underlying reasons is that the naked person should be punished through criminal law. In contrast the responses of the people in the constraint condition should represent the application of the exclusionary model. According to this model, the naked person's behaviour should not be criminally punished,66 yet, on the face of it, it appears that well-motivated persons, aware of the task of exclusion, are unable to neutralise their moral deliberation.

• 67 Sood 2015: 1563 f.

76If the experiments seen in the previous section prove the experimenters’ hypothesis, then we have reason to believe that there is such a thing as involuntary disobedience to law. As Sood nicely puts it, elaborating on a position expressed by Robert Cover, a judge who is caught between law and morality seems to have only four options: to apply the law against morality; to follow morality rather than the law; to resign; and to cheat, that is to resolve the conflict between law and morality by resorting to an interpretation of the law she does not believe. To these options, Sood notes, we must add a fifth in which “judges may unintentionally construe facts and apply law in a way that preserves the appearance — not only to others but also to themselves — of conforming to both the legal doctrine and their own intuitions about justice”.67

• 68 Wistrich, Guthrie & Rachlinski 2005: 1259, 1322.

78First of all, it must be said that Wistrich, Guthrie, and Rachlisnki themselves, in the study reported above, note how in several cases — especially those involving rights protected by the U.S. Constitution — judges effectively neutralise the impact of inadmissible information on their reasoning.68 The extent of the failure of the exclusion task is therefore still to be explored.

• 69 Bublitz 2020: 15 f.
• 70 According to Bublitz, legal errors must be divided between explicit errors, which can be found in t (...)

79Second, it should be noted — for the moment only en passant — that the category of involuntary disobedience resolutely clashes with a certain way of understanding the law. On some views, legal provisions are indeterminate: “Judging requires applying loosely defined concepts or broad standards to a concrete case, tailored to its circumstances through interpretation”.69 Raz’s exclusionary model often seems tailored to cases that do not raise great interpretative problems. If we take into account legal indeterminacy and interpretative problems, it is no longer obvious on what basis it is possible to censure the application of the rape-shield statute or the manipulation of the harm principle as cases of involuntary disobedience to law. Consider, in the experiment about the harm principle, the thesis according to which when people criminalize the conduct of going naked to the supermarket they are manipulating the concept of harm. Against this idea someone may retort by observing that reasonable people may interpret this conduct as a source of harms, because, for example, “children may see it and this will hurt them”, or “this is offensive behaviour, and offensive behaviour is harmful” or “such behaviour may stir people’s sexual desires in a way that could cause harm”. This does not mean that the authors of these studies have neglected the entire issue outright. In the experiment about the manipulation of the harm principle the first part of the experiment is concerned about the individuation of a conduct that people spontaneously take to be harmless. This part aims precisely to counter the objection according to which people simply disagree about what conduct is harmful. Thanks to this part of the experiment, it is plausible to believe that people formulated an exclusion task and then failed to perform it. Still, it remains possible that going to the supermarket naked may be sensibly interpreted as harmful conduct because children may be hurt, because offensive behaviour may be harmful, and so on. How exactly the experimental results in favour of unintentional disobedience can be reconciled with the idea of legal indeterminacy and the problem of interpretation is something that requires further elaboration.70

• 71 Gur 2018: 129 f.

80However, even with these limitations, this paper develops the integration process between Raz's theory of authority and cognitive science studies. In this regard, it should be mentioned that Gur's analysis of biases is not limited to the weighing model. He believes that the exclusionary model also fails to counteract biases to a successful degree. For my analysis to contribute to the debate I must explain where the problem of mental contamination differs from Gur’s critique of the exclusionary model. Gur’s general idea is that the exclusionary model “better enables law to counteract those biases than the weighing model does, since the latter, unlike the former, invariably requires subjects to act through all-things-considered balancing exercises”.71 The reason why Gur believes that the exclusionary model doesn’t counteract biases to a successful degree revolves around the piecemeal structure of Raz’s theory of authority, which is derived by the so called “normal justification thesis”, according to which

• 72 Gur 2018: 128.
• 73 About this second problem Gur recognizes that Raz may retort by saying that in assessing the superi (...)

81This means that to test whether one person has authority over another we need to test the advantage of obedience on an individual basis and in a domain-specific manner. Following this idea, for example, the expert pharmacologist is not subject to the authority of laws about the safety of drugs. But, as Gur notes, if the scope of legitimate authorities is so patchy, people willing to follow the exclusionary model are asked to act “under unfavourable conditions that involve potential exposures to biases”.72 First, their assessment of superiority over the person claiming authority is subject to bias, as is their balancing of reasons. Second, even in the cases in which the subject is really superior as a matter of expertise than the person claiming authority, she will be affected by other biases not linked to epistemic superiority — biases that have “little to do with the lack of information”.73

• 74 Raz 1986: 39.
• 75 Gur 2018: ch. VII.
• 76 Thaler & Sunstein 2008.
• 77 Elster 1984 [1979]: 49; the concept of pre-commitment has already been explored in its connection w (...)

86Finally, I would like to draw attention to an answer or, perhaps better, a possibility of metamorphosis that remains open for the exclusionary model. As mentioned above, Raz’s said that “possibly it [reflection on the merits of actions] could be prohibited by a special directive to that effect.”74 Studies on mental contamination show that we are sometimes unable to counter biases despite our awareness and motivation. If so, our mental power to exclude (what we regard as) first-order reasons defeated by exclusionary reasons is fallible. So, how can we counter biases more efficiently? There are several possibilities, like the cultivation of the disposition to obey the law (this is the core idea of Gur’s dispositional model75), or investments in choice architecture76 leading to a real compartmentalization of information (measures that are already common in the adversarial system in criminal law). To provisionally summarize all these measures under one label, we may say that a possible solution to the problem of mental contamination is to exploit pre-commitment strategies, which in Elster’s terms may be defined as any decision adopted at time T1 that increases the probability that one will carry out another decision at time T2.77

• 78 Vassiliou 2022: 945 ff.; Chang 2016.

88A distinction can be made between models about practical reasons and models about practical reasoning.78 The former are concerned with answering questions about what reasons for action exist and how they interact; the latter are concerned with prescribing decision-making processes. As far as we interpret the exclusionary model as prescribing a specific mental performance — the exclusion task — we see it as a model about practical reasoning. But it would cease to be a model of practical reasoning and become a theory of practical reasons if we took the model as the assertion that there are sometimes reasons not to act on certain reasons, without any specific commitment to a specific decision-making process to arrive at the correct practical solution. A theory which, in other words, indicates just the goal of the exclusion of reasons, legitimising a wide variety of strategies that could lead to this result.