From Angels to Humans: Law, Coercion, and the Society of Angels Thought Experiment

Whether legal systems are necessarily coercive raises normative concerns. Coercion carries a presumption of illegitimacy and a special justificatory burden. If legal systems are necessarily coercive, coerciveness necessarily taints our legal institutions. Traditionally, legal systems have been regarded as contingently coercive. This view is mainly supported by the society of angels thought experiment. For the past few years, however, this traditional view has been under attack. Critics have challenged the reliability of the thought experiment and have urged us to centre the discussion on typical legal systems: legal systems made by humans to address human needs. Once we do so – they claim – we would inevitably reject the traditional view. This paper argues that the critics are wrong. After discussing key features of the society of angels thought experiment and responding to objections, it is argued that even typical legal systems are contingently coercive. Coerciveness is a feature that our legal systems can and should strive to get rid of.

FROM ANGELS TO HUMANS: LAW, COERCION, AND THE SOCIETY OF ANGELS THOUGHT EXPERIMENT (Accepted 16 November 2020) ABSTRACT. Whether legal systems are necessarily coercive raises normative concerns. Coercion carries a presumption of illegitimacy and a special justificatory burden. If legal systems are necessarily coercive, coerciveness necessarily taints our legal institutions. Traditionally, legal systems have been regarded as contingently coercive. This view is mainly supported by the society of angels thought experiment. For the past few years, however, this traditional view has been under attack. Critics have challenged the reliability of the thought experiment and have urged us to centre the discussion on typical legal systems: legal systems made by humans to address human needs. Once we do so -they claim -we would inevitably reject the traditional view. This paper argues that the critics are wrong. After discussing key features of the society of angels thought experiment and responding to objections, it is argued that even typical legal systems are contingently coercive. Coerciveness is a feature that our legal systems can and should strive to get rid of.

I. INTRODUCTION
Whether legal systems are necessarily coercive raises normative concerns. Coercion carries a presumption of illegitimacy and a special justificatory burden. If legal systems are necessarily coercive, the 'taint of coerciveness' 1 befalls our legal institutions wherever and whichever way they occur. Traditionally, legal philosophers see coerciveness as a mere contingent feature of legal systems. For them, legal systems are not the type of institution that necessarily carries a presumption of illegitimacy or has the everlasting need to discharge this presumption with a special justification. For the past few years, however, this traditional view has been under attack. The challenge is motivated by methodological doubts about the reliability of the main thought experiment that supports the traditional view: the society of angels thought experiment. Critics have argued that we should not make claims about legal systems based on scenarios depicting institutions designed to deal with angelic needs; we should focus instead on typical legal systems: legal systems made by humans to address human needs. 2 Once we do so -they claim -we should abandon the traditional view and acknowledge that typical legal systems are necessarily plagued by coerciveness and its stains. 3 This paper argues that the critics are wrong. Sections II and III discuss the society of angels thought experiment's main features, its role, and respond to two objections against its reliability. Section IV moves away from the society of angels and presents two new arguments for the traditional view. According to these arguments, even typical legal systems are not necessarily coercive. Coerciveness is a contingent feature of our legal systems, one that we can and should strive to get rid of.

II. THE SOCIETY OF ANGELS
Thought experiments featuring angels have for long found their place in philosophy. Medieval philosophers, for example, have used thought experiments involving angels to discuss the possibility of having cognitive access to the essence of things and to clarify the relation between thought and language. 4 In legal and political philosophy, scenarios involving angels or other morally-driven subjects appear occasionally, at least since Aristotle. 5 Within this context, perhaps the most famous occurrence is Madison's claim that 'if men were angels, no government would be necessary'. 6 Closer to our time, what became known as the 'society of angels thought experiment' has received a fair amount of attention in contemporary analytical jurisprudence. Several legal philosophersincluding prominent figures such as Hart,7 Raz,8 Finnis,9 and Gardner 10 -have presented variations of this thought experiment while discussing the role of coercion in law. In this discussion, the thought experiment is supposed to both work as a counterexample to the claim that legal systems are necessarily coercive and to provide support for the view according to which non-coercive legal systems are possible. Here is how the thought experiment goes: (Society of Angels): Imagine a society of morally perfect angels. Despite being morally perfect, the angels in this society can still make use of a legal system to coordinate numerous aspects of their public lives and resolve whatever disputes may arise. Nonetheless, the angels in this society need not be coerced to cooperate with one another and to achieve their social goals. Hence, the angelic legal system need not resort to coercion to properly work. 11 It is worth noting that the society of angels thought experiment is not itself an argument for the claim that non-coercive legal systems are possible; 12 it is simply a description of a scenario where a noncoercive legal system exists. The scenario, of course, is used in an argumentative context. Namely, it is used to provide a reason in support of the claim that a non-coercive legal system is possible.
Yet, it is not entirely clear what it means to say that the thought experiment provides a reason in support of the claim that a noncoercive legal system is possible. There are least two worries.  (Oxford: Oxford University Press, 2006), pp. 208-209. 11 Note that the background discussion in which the thought experiment appears concerns the need of coercion to ensure the efficacy (and existence) of legal systems. There are, however, different discussions about the relationship between legal systems and coercion. One such discussion concerns whether the authorisation of the use of coercion is necessary for a legal system to create legal obligations. The society of angels thought experiment does not (directly) bear on this discussion. I thank one of the reviewers for pressing me on this point. 12  first concerns the relevant meaning of 'possible'. The second relates to our justification for believing that the scenario depicted by this thought experiment is possible in the relevant sense. I deal with the first worry in the following subsection and with the second in sequence.

A. What Kind of Possibility?
Joseph Raz is probably the most well-known user of a variation of the society of angels thought experiment. He claims that '[e]ven a society of angels may have a need for legislative authority to ensure co-ordination' 13 and that courts would be useful in such scenario. 14 Raz's remarks on the thought experiment can help us find the relevant notion of possibility for the discussion.
Raz asks rhetorically whether a non-coercive legal system is possible. But his answer is prefaced with a caveat: the relevant notion of possibility at stake is not human possibility. For Raz, and for many others, a non-coercive legal system is 'humanly impossible'. 15 With this Raz most likely means that a non-coercive human legal system is nomologically impossible; 16 it is impossible given the biological and psychological dispositions of human beings.
Though humanly impossible -Raz continues -a non-coercive legal system is 'logically possible'. 17 It is so because, according to Raz, it is plausible to conceive 'other rational beings who may be subject to law, who have, and who would acknowledge that they have, more than enough reasons to obey the law regardless of sanctions'. 18 From this, it can be inferred that Raz presents the thought experiment as a reason for the logical possibility of a noncoercive legal system.
Even though Raz explicitly states that the relevant notion of possibility is logical, we should not interpret his statement literally. A given proposition is logically possible if, and only if, it is not a logical falsity. Thus, we can identify logical possibilities and impossibilities just by looking at the logical form of propositions. That an object is square and not square at the same time is a logical impossibility. But that Usain Bolt is a cake is a logical possibility -the proposition 'Usain Bolt is a cake' is not a logical falsity. 19 As we can see, logical possibilities are too permissive. For that reason, despite being true, the claim that a non-coercive legal system is logically possible is trivial and probably uncontentious. There are, therefore, good reasons to neither interpret Raz's claim nor the discussion revolving around the claim that legal systems are necessarily coercive as being about the logical possibility of a non-coercive legal system. This possibility is too broad to be of any theoretical significance.
The best way to interpret Raz's and similar claims -I'll argue -is interpreting them as claims of metaphysical possibility. 20 Something is 17 Raz (n 8) 158. 18 Ibid., 159. Note that, strictly speaking, Raz is imagining a scenario featuring a sanctionless legal system. There are, however, good reasons to think that sanctions are neither necessary nor sufficient for a legal system to be coercive. See Here I'll bracket these worries and treat Raz's thought experiment as if it was intended to support the claim that a non-coercive legal system is possible. 19 Take C as the property of being a cake, and a as the arbitrary name that refers to Usain Bolt. The proposition 'Usain Bolt is a cake' can be formalized as follows Ca. This proposition is false. But it is not false as a matter of logic: the logical form is not a logical falsity (e.g., a contradiction). 20  metaphysically possible if, and only if, it is consistent with the most fundamental metaphysical principles and categories that structure reality. 21 When we ask, for example, if it is metaphysically possible for Usain Bolt to be a cake, we want to know if, given how reality is fundamentally arranged, a circumstance where Usain Bolt is a cake is genuinely a case where Usain Bolt, and not someone or something else, obtains. That is, whether whatever it takes for being Usain Bolt is compatible with what it takes to be a cake.
Metaphysical possibility differs from both nomological and logical possibilities. It is broader than nomological possibility, but not as broad as logical possibility. 22 That is, there are metaphysical possibilities which aren't nomological possibilities, and logical possibilities which aren't metaphysical possibilities. For example, it is metaphysically possible for Usain Bolt to run faster than the fastest of cheetahs even though it is nomologically impossible. 23 And it is metaphysically impossible for Usain Bolt to be a cake, although it is logically possible. It is difficult to find uncontroversial examples of mere metaphysical possibility. Despite this difficulty, intuitive examples such as the one where Usain Bolt is a cake help us to make sense of the notion.
We've seen that construing the claim that a non-coercive legal system is possible as a claim about logical possibility makes the relevant discussion trivial and uninteresting. That doesn't happen when we construe it as a claim about metaphysical possibility. In fact, doing so helps us better understand the broader discussion in which the debate about whether legal systems are necessarily coercive is a part.
The debate over whether legal systems are necessarily coercive is often assumed to be part of a broader inquiry about the nature of law. 24  something; 'questions about the nature of some thing are paradigmatically metaphysical questions'. 25 They are, in other words, questions about the object-level: questions about things, as they occur in reality; not questions about how we represent or conceptualise things, questions about the representational-level. Now, if we see -as I think we should -the debate about legal systems being necessarily coercive as a debate about the metaphysics of legal systems, it becomes clear why the notion of metaphysical possibility is the relevant one. If we are interested in the metaphysics of legal systems -in the nature of law, as it were -we need to find out if the kind of thing we happen to call 'legal systems' could have been noncoercive and still remain the same kind of thing.
Nevertheless, not all legal philosophers who discuss the nature of law in general, and the necessity of coercion to legal systems in particular, see themselves as engaged in metaphysical debates. Quite the opposite. As Plunkett and Shapiro have recently suggested, and as confirmed by many works in the methodology of legal philosophy, 26 'many [legal] philosophers (…) harbor deep suspicion about metaphysics and don't spend much (if any) time working on it.' 27 Instead, many legal philosophers prefer to see both the debate about the nature of law, and the debate about the necessity of coercion to legal systems, as conceptual debates. They are interested, for example, in investigating if our concept of legal system, as we have it, applies to a non-coercive institution such as the one imagined in the society of angels thought experiment. 28 Thus, for those philosophers, the relevant sense of possibility for the discussion is not metaphysical; it is conceptual. The notion of conceptual possibility is elusive and is used by philosophers in different ways. 29 Most common among legal philosophers is the following notion of conceptual possibility: something is conceptually possible if, and only if, it is consistent with our conceptual framework; with how we currently and ordinarily represent the world. 30 Following this interpretation, a non-coercive legal system is conceptually possible, if, and only if, our concept of legal system extends to a non-coercive system.
From what I've said above, we can see that the relevant sense of possibility for the discussion about the society of angels thought experiment hangs on a broader debate about the proper object of a theory about the nature of law. Adjudicating between these two conceptions of the object of a theory about the nature of law (conceptual and metaphysical) is beyond the scope of this paper. However, let me clarify why it strikes me as more plausible, or at least more interesting, to see metaphysical possibility as the relevant sense of possibility for the society of angels thought experiment. 31 First, it is pertinent to point out a mistake in a reason for preferring the conceptual approach sometimes found in the works of legal philosophers. Some have favoured the conceptual approach in virtue of thinking that it fares better with two interrelated assumptions about legal systems they consider sound: the assumption that 29 Sometimes it is used as synonymous to metaphysical possibility, sometimes it is used as a subtype of logical possibility, and sometimes it is used as cohering with our ordinary concepts. See, Lowe (n 21) 13-14. 30 The way legal philosophers generally understand conceptual possibility is heavily influenced by Frank Jackson, From Metaphysics to Ethics: A Defence of Conceptual Analysis (Oxford: Oxford University Press, 1998), p. 31; Frank Jackson, ''Thought Experiments and Possibilities'', Analysis 69(1) (2009): pp. 100-109. It is worth mentioning that Frank Jackson has never elaborated on the relevant theory of concept acquisition and formation that the notion of conceptual possibility requires. This is likely to generate problems. For an argument challenging Jackson's notion of conceptual possibility on these grounds, see Herman Cappelen, Philosophy Without Intuitions (Oxford: Oxford University Press, 2012), p. 216. 31 Philosophers who favour viewing philosophical debates of the sort here considered as metaphysical tend to have a deep suspicion about the plausibility and philosophical relevance of conceptual truths. Some go as far as rejecting the notions of conceptual possibility and necessity altogether. legal systems are social constructs and the assumption that social constructs are nothing but what people collectively think and speak about them. 32 The thought is that if legal systems, as social constructs, are entirely constituted by mind dependent facts about our linguistic or conceptual practices and behaviour, then all the truths about legal systems, including truths about what is necessary and possible about legal systems, would be found in our thought and talk about legal systems. If that is the case -the argument goes -then relying on the notion of conceptual possibility seems to make more sense.
This is a non-sequitur. From the assumption that legal systems are entirely constituted by the sort of mind dependent facts mentioned (an assumption that is much more controversial than some legal philosophers would have us believe), 33 it doesn't follow that the notion of conceptual possibility is better suited than metaphysical possibility. Both notions would elicit exactly the same results in this scenario: they would be explanatorily equivalent. This is because, in this scenario, whatever is metaphysically true of legal systems is coextensive with what is conceptually true of it: legal systems are the sort of thing that is entirely constituted by the way people think and speak about it.
Ironically, the very move that was supposed to get rid of metaphysical talk about legal systems ends up with a notion of conceptual possibility that is explanatorily on par with the notion of metaphysical possibility. What is more, the very assumption that legal philosophers rely on to establish the priority of conceptual modality i.e., the assumption that legal systems are entirely constituted by facts about our thought and talk -is metaphysical. To substantiate it, legal philosophers would have no other option but to engage in the very kind of metaphysical discussion they wanted to eliminate with the adoption of the conceptual approach. All suggests that the aversion some legal philosophers have to metaphysical discussions and the notion of metaphysical possibility is uncalled for. If that is the case, there seems to be no clear advantage in preferring the conceptual approach. But there are costs.
The most obvious one is being revisionist about a methodology that most philosophers arguably rely on when they make other types of claims about legal systems. For example, philosophers have claimed that typical legal systems are pervasively coercive, 34 that typical legal systems coordinate behaviour, 35 that some of its norms are unjust, etc. These claims are meant to be taken as claims about typical legal systems; as claims directly about the kind of institutions that exist and apply to us. These aren't claims about how we think or speak about typical legal systems; claims about our representation of typical legal systems. It strikes me that this is also how most people would see these claims.
If this is so, then why assume that things should be radically different when we discuss features that are necessary to legal systems? More to the point, why assume that when discussing the necessity of coercion to legal systems we should no longer see this as a discussion about legal systems but instead as a discussion about our representation of legal systems?
These worries disappear if we adopt the metaphysical approach. Adopting the metaphysical approach has the advantage of making discussions about necessary features of legal systems methodologically continuous to the kind of discussion philosophers usually have when discussing features of typical legal systems. In other words, the metaphysical approach doesn't call for a reorientation in the methodology most philosophers already rely on to discuss more common questions such as 'Are typical legal systems pervasively coercive?'.
The society of angels thought experiment gives us a reason for believing in the possibility of a non-coercive legal system. If I'm right about what I've said in this subsection -and to make modal discussions about legal systems methodologically continuous to other common philosophical discussions about them -we should treat the society of angels thought experiment as giving us a reason to believe in the metaphysical possibility of a non-coercive legal system. Yet, nothing I mentioned in this subsection clarifies how we can be justified in believing that the scenario depicted in the society of angels thought experiment is in fact metaphysically possible. This is the subject of the next subsection.

B. Knowing Possible Legal Systems
Knowledge of metaphysical possibility is a central topic in modal epistemology. Some philosophers argue that we know what is metaphysically possible through our ability to conceive scenarios. According to this account, if something is conceivable, it is metaphysically possible: conceivability entails metaphysical possibility. 36 Others reject the conceivability account and defend that our knowledge of metaphysical possibilities is just a special case of our knowledge of counterfactuals. 37 According to this account, the way in which we know if something is metaphysically possible is not different from the way in which we know the truth of common counterfactual propositionse.g., 'If it had rained, the game would have been cancelled'.
Adjudicating between different approaches to the epistemology of modality is not something I intend to do in this paper. For the present purposes, I'll only show how we can (at least) justifiably state that a non-coercive legal system, as depicted in the society of angels thought experiment, is metaphysically possible. I'll do so while trying not to take a stake in deep controversies in modal epistemology.
My claim is that background knowledge 38 and ordinary experience can help us to justify the belief that a non-coercive legal system (as depicted by the thought experiment) is possible. Here is how it can help. First, background knowledge and experience set an initial baseline for what is metaphysically possible: whatever is actual is metaphysically possible. If a thought experiment has features that are actual, we can infer that at least those features are metaphysically possible. 36 Stephen Second, background knowledge and experience can be used as defeasible guides to analyse the plausibility of metaphysical possibility claims. It is plausible to assume, based on our background knowledge, that the English legal system could have had fewer rules than it actually does. But it is not plausible to assume that it could have had no rules at all. That would clash with our experience and background knowledge about the role and the structure of legal systems. Yet, it is a defeasible guide because proximity to what we know cannot be a conclusive reason; we are fallible and have limited knowledge. Our experience and background knowledge may be proven wrong by stronger considerations, such as a sound theory about the nature of legal systems. But in the absence of such considerations, and given all we currently know and have experienced about legal systems, we wouldn't be justified in believing that a ruleless legal system is metaphysically possible.
Having said that, here are some considerations that illustrate how background knowledge and ordinary experience can help us to justify the claim that a non-coercive legal system, as depicted in the society of angels thought experiment, is metaphysically possible.
First, we know that the degree of coerciveness of legal systems can vary from system to system and within the same system at different time frames. These variations can be drastic (as when a military coup or a radical change in the political orientation of a government takes place) or mild (as when the legal system abolishes a particular kind of punishment). This means that we know that legal systems can adjust (and usually tend to adjust) the use and the availability of coercive mechanisms to respond to particular circumstances and needs. Which means that it is metaphysically possible for a legal system to drastically reduce its degree of coerciveness when coercion proves unnecessary for that legal system to work (or for any other reason).
We also have background knowledge about what sorts of activities fall within the scope of legal systems. We know that actual legal systems don't merely punish, deter crimes, or compel people to do things against their will. As pointed out by philosophers and social scientists, a significant part of the activities performed by legal sys-tems have little to do with coercion. 39 Legal systems play an important role in solving coordination problems and in disseminating information about people's empirical and normative expectations. 40 Legal systems regulate contracts, wills and marriage, separate private from public, design tax schemes, regulate the allocation of public goods, and do things like the determination of frequencies of radio and TV, the form of traffic signs, the processes involved in elections and in the transition of governments, assign roles and role obligations, and so on. We know, therefore, that non-coercive activities are a non-trivial part of what legal systems are used for.
A feature of the thought experiment that may give rise to some doubts is moral perfection. 41 We need not, however, worry too much about it. A detailed account of moral perfection is unnecessary for the thought experiment to run. All that is needed is that moral perfection entails that angels would cooperate with one another when cooperation prevents the occurrence of morally bad outcomes and when cooperation helps angels to achieve morally good outcomes. This can be grounded on our background knowledge and ordinary experience: morally good people are cooperative when cooperation helps them in achieving a morally better state of affairs; a fortiori, morally perfect people would also be cooperative in the same circumstances. Now, take both the fact that angels are cooperative and the fact that legal systems tend to reduce their degree of coerciveness where coercion is proven unnecessary for the system to work. From these facts, it is plausible to conclude that a normative institutional system in a society of angels could have a null degree of coercion. The difficult part, however, is moving from this conclusion to the claim that the system present in the society of angels is indeed a legal system.
What can grant some plausibility to this claim is the large overlap between the social needs that the system in the society angels helps to fulfil (and is purported to fulfil) and the social needs actual legal systems do fulfil. A society of angels might still have the need to create rules to allocate property, to regulate contracts, wills, taxation, to solve small and large coordination problems related to public goods, political processes, the organization of common space (including the organisation of traffic, zoning, signals, etc), assigning roles, allocating risk, settling disputes, and many other non-trivial activities that fall within the scope of the activities performed by actual legal systems.
Given the proximity between the institutional system present in the society of angels and our legal systems, there is a reason to believe that angels have a legal system, and not a different kind of institution. The thought experiment, however, shouldn't be taken as a silver bullet to decide the question of whether non-coercive legal systems are metaphysically possible. The reason one can extract from the thought experiment is defeasible and it must be weighed against countervailing considerations.
Insofar as it is used to provide a defeasible reason for the metaphysical possibility of a non-coercive legal system, the angelic thought experiment is in good order. Some philosophers, however, are suspicious about this thought experiment and have attempted to undermine it. The next section discusses the main objections presented against the angelic thought experiment.

III. ANGELS ON TRIAL
Some philosophers have challenged the use of the society of angels thought experiment. In what follows, I discuss two main objections against it.

A. The Under-Specification Objection
The objection can be formulated as follows: The Under-specification Objection: The angelic thought experiment lacks crucial details. Therefore, it cannot provide a good reason for the metaphysical possibility of a non-coercive legal system.
Andrei Marmor is one of its proponents. He argues that the society of angels thought experiment is 'inconclusive' given that 'it is far from clear how the rationality of the imaginary angels in our thought experiment should be defined'. 42 Along the same line, Dan Priel argues that, 'it is not clear what we could learn from such an otherworldly example as its terms are so unspecified'. 43 One could wonder about what kind of details the thought experiment must lack to be considered inconclusive. It is difficult to specify all of them. Nonetheless, it is possible to offer general guidelines by looking at what we know about the reasons why legal systems coerce citizens. For Marmor, for instance, the thought experiment is inconclusive if it is not clear whether angels would face prisoner dilemma situations. If they do, then -Marmor claimsangelic legal systems must rely on coercion in order to overcome 'problems of opportunism'. 44 The objection doesn't work. At least three replies are possible. First, the angelic system could simply issue a law determining that angels ought to cooperate whenever a prisoner dilemma situation occurs. Since angels would obey their laws, that would be enough for solving prisoner dilemma situations. Second, we can dispute Marmor's assumption that the angelic system must resort to coercion to solve prisoner dilemma situations. That assumption disregards alternative ways to solve prisoner dilemma situations and problems of opportunism in general; ways that may be much more coherent with the moral character of angels and perhaps even more effective to solve these dilemmas. The angelic system could, for instance, have laws providing a reward for those who cooperate in prisoner dilemma situations, or laws that broadcast angels' normative expectations against opportunism, or an intricate system of nudges. There is no reason to think that coercion is necessary to solve prisoner dilemma situations. This is not a prerogative of a society of angels; the same considerations apply to human societies. 45 The third reply to Marmor's objection is to challenge the claim that the thought experiment is unspecified in the aspect that he pointed out. Details such as the occurrence of prisoner dilemma situations can be inferred from what has been stipulated about angels' moral character. As mentioned earlier, that angels are cooper- ative can be inferred from the assumption that they are morally perfect. Cooperation generates the best outcome for both parties in a prisoner dilemma. Since the dilemma only occurs when there is doubt about whether the agents in the dilemma are cooperative and angels would have a strong expectation (perhaps even knowledge) that other angels are cooperative, no such dilemma would occur in a society of angels.
Yet, even if we concede that the objection is appealing, there is no need to worry too much about it. If sound, the objection would undermine just a version of the thought experiment. There are still other versions where these details are specified. If the problem of the thought experiment is just a matter of under-specification, one could simply present an alternative, and more detailed, version of it.

B. The Unreliability Objection
The second objection I consider is this: The Unreliability Objection: The reason we extract from the thought experiment is not reliable. Therefore, it does not adequately support the claim that a non-coercive legal system is possible.
The general idea here is that thought experiments depicting a society of angels or morally perfect individuals are too far removed from our ordinary perception of things. Therefore, we cannot trust our background knowledge about them. Marmor advances this objection when he claims that '[t]he further our angels are removed from the human condition, the less clear it becomes what we are supposed to learn from the example'. 46 Dan Priel, in turn, goes further and argues that '[s]ince we have not encountered a society of angels and have not had to think about whether what they have is 'really' law, the application of our concept of law to such societies is indeterminate'. 47 Priel's claim that we have never encountered a society of angels raises an initial worry. It is not clear if Priel is assuming that we must have empirical or perceptual knowledge about a society of angels in order to reliably draw conclusions about it. If that is the way we should interpret his claim, his objection fails.
The idea that we need empirical or perceptual knowledge about a society of angels in order to draw reliable conclusions about it strikes me as an odd one. It is unclear if Priel thinks that this is a general requirement about knowledge. If so, we should discard any use of thought experiments and counterexamples that cannot be empirically replicated. Not only that: if there is a general requirement according to which we can only rely on what can be empirically or perceptually verified, then we shouldn't rely on the very claim that we should only rely on what can be empirically or perceptually verified. If, on the other hand, it should be taken as a requirement applicable just to the society of angels thought experiment, it would need further justification.
But perhaps this was not the reading intended by Priel. Perhaps what he meant was something similar to Marmor's objection. According to this reading, the fact that the scenario depicted by the angelic thought experiment is too alien from all we know constrains our ability to rely on our background knowledge, which is based on human legal systems. This reading is less radical and more plausible.
The general point of this objection is correct: the more distant the possibilities considered, the less reliable our judgements about them are. We have limited cognitive capacities and it is difficult to be precise about alien scenarios. At the same time, however, this is not a legitimate concern for the society of angels thought experiment. What we need to stipulate in the society of angels' scenario is clear enough: it is a society solely inhabited by creatures that resemble us in every respect except from being much more cooperative and lawabiding. As I've claimed in the previous section, our background knowledge and ordinary experience can help in this scenario. From the stipulated features, we can extract angels' general behaviour and their social needs.
Of course, some of the drawn implications can be disputed. From the fact that angels are morally perfect one could conclude that there would be no need for courts to adjudicate disputes about facts. (But, even if this is true, couldn't angels have institutions they don't strictly need? Maybe having a legal system is just more convenient to them). This just reinforces the point I highlighted in the previous section: the thought experiment is not supposed to be taken as a silver bullet that decides the question. It is merely used as a defeasible reason for the metaphysical possibility of a non-coercive legal system.
But doesn't the kind of disagreement just mentioned reinforce the unreliability objection? After all, the fact that people disagree about the implications drawn from the thought experiment may occur precisely because of the removed nature of the scenario described in the society of angels. That is too hasty. There is no good reason to think that the disagreement occurs mainly in virtue of some peculiarity of the thought experiment. Rather, it occurs in virtue of the nature of the implications drawn from the scenario. The implications drawn are philosophical implications; they aren't disputes about empirical facts. As such, there is no guarantee that were the thought experiment less removed from our common perception of things, there would be no (or fewer) disagreements about the conclusions drawn. The vast number of controversies stemming from less removed thought experiments such as Gettier cases -experiments that can be easily empirically replicated -reinforce that. 48 The objection against drawing reliable evidence from the angelic thought experiment can be met. A caveat, however, is in order. Despite what I've said in its defence, the angelic scenario is just too fictitious for some philosophers and lawyers to take it seriously. Even if they eventually concede that the experiment supports the metaphysical possibility of non-coercive legal systems, they may still deem the discussion around this kind of possibility esoteric and uninteresting. No matter how plausible or how grounded in our background knowledge the angelic thought experiment is, critics will still insist that it won't help us draw interesting conclusions about the coerciveness of typical legal systems; namely, it won't tell us whether a legal systems for humans could possibly exist without coercion. That -and not whether angels can have a non-coercive legal system -is what critics want to know. 49 This is a fair point. 50 Thus, if we want to make a more convincing and perhaps more 48 Edmund L Gettier, ''Is Justified True Belief Knowledge?'', Analysis 23(6) (1963): pp. 121-123. 49 In other words, they seem to be concerned with the nomological possibility of human legal systems being non-coercive. 50 Though it does not undermine the conclusion drawn from the society of angels thought experiment: that non-coercive legal systems are metaphysically possible.
interesting case for the claim that legal systems are not necessarily coercive, we must branch away from the society of angels and argue that typical legal systems, legal systems created by humans to address human needs, are not necessarily coercive. That is what I will do next.
IV. BY HUMANS, FOR HUMANS Some philosophers are solely concerned with features of typical legal systems. For them, the society of angels is just an old trinket of analytical jurisprudence that has no bearing on our understanding of typical legal systems. When we look around -these philosophers would say -we see that coercion is not only a salient feature of our legal systems, but that typical legal systems cannot exist and work without coercion. Here I'll present two arguments against this view; arguments that show that that even typical legal systems are contingently coercive. 51

A. The Argument from the Wounded State
Typical legal systems -systems made by humans for humans -back their legal mandates with sanctions and other enforcement mechanisms. In fact, sanctions and other enforcement mechanisms are so central to the role typical legal systems play in our societies that scenarios where normative systems lack such mechanisms invite scepticism towards their status as legal system. That is another reason why some are unwilling to accept arguments for the contingency of coerciveness that rely on the society of angels thought experiment.
But if our goal is to argue for the idea that legal systems (typical and atypical) are contingently coercive we need not consider a scenario featuring a system devoid of sanctions and other enforcement mechanisms. As many legal philosophers have pointed out, the coerciveness of legal systems is not to be conflated with the presence of sanctions and enforcement mechanisms. 52 To be coercive, enforce-51 The arguments will show that a non-coercive typical legal system is nomologically possible. A fortiori, they stablish that non-coercive legal systems are metaphysically possible, which is the same conclusion supported by the society of angels thought experiment. 52 See Oberdiek (n 18); Lamond (n 18); Yankah (n 18); Green (n 28). ment mechanisms must generate particular effects. There is of course no agreement about the exact effects these mechanisms must generate in order to be coercive. 53 Despite the lack of agreement on the details, most philosophers accept that enforcement mechanisms are coercive only if they make non-compliance with legal mandates unreasonable in some non-trivial sense (to be specified by an account of coercion). 54 That coerciveness is not tantamount to the presence of sanctions and enforcement mechanisms and that a legal system is coercive only when the enforcement mechanisms render non-compliance with legal mandates unreasonable in some non-trivial sense is all an argument for the contingency of coerciveness in legal systems needs. To make such an argument, I will rely on the following thought experiment presented by Andrea Sangiovanni (though in connection to a different discussion): Imagine an internally just state. Let us now suppose that all local means of law enforcement-police, army, and any potential replacements-are temporarily disarmed and disabled by a terrorist attack. Suppose further that this condition continues for several years. Crime rates increase, compliance with the laws decreases, but society does not dissolve at a stroke into a war of all against all. Citizens generally feel a sense of solidarity in the wake of the attack, and a desire to maintain public order and decency despite the private advantages they could gain through disobedience and noncompliance; this sense of solidarity is common knowledge and sufficient to provide assurance that people will (generally) continue to comply with the law. The laws still earn most people's respect: the state continues to provide the services it always has; the legislature meets regularly; laws are debated and passed; contracts and wills drawn up; property transferred in accordance with law; disputes settled through legal arbitration, and so on. 55 The system in the scenario above no longer has resources to perform law-enforcement functions. Though nothing is said about the presence of sanctions and enforcement mechanisms, we can assume that they still exist but are not enforced. The inhabitants of the wounded state know that authorities will not enforce legal mandates: state agents will not hurt, force, detain, search, or kill 53 Some believe that they must pose 'major restrictions on fundamental human freedoms' (e.g., Oberdiek (n 18) 82.), others that they must compel (e.g., Yankah (n 18) 1217., yet others that they must give a sufficient reason for performing or not performing a particular course of action (e.g., Lamond (n 18) 52.). 54 If individuals have reasons or motives to act according to their preferences or desires, then any demand or act that makes one's choices or courses of action less eligible would be-in a trivial sense-unreasonable. The qualifier ''in some non-trivial sense'' is therefore added to circumscribe the predicate ''coercive'' to more robust interferences with people's choices or acts. In order to remain as ecumenical as possible, I will not presuppose any account of coercion (or ''unreasonableness''). An intuitive, pre-theoretical, understanding of what counts as ''serious'' or ''robust'' interferences is enough for my purposes. 55  citizens in order to safeguard citizens' rights and liberties or to ensure that citizens perform their duties. As a result, the inhabitants of the wounded state can infer that non-compliance with legal mandates is not unreasonable; the increase of non-compliance and criminal activity corroborates that. Coercion, therefore, has disappeared from the scene. Yet, as described in the scenario, most citizens still comply with the mandates of this system, and -except for law-enforcement -the system still successfully performs all the functions it used to perform before being struck by the terrorist attack. 56 The scenario described in Sangiovanni's thought experiment is much closer to what we are familiar with; it describes a system designed to work in a human society. The assumptions that the thought experiment at hand makes are firmly grounded in our background knowledge about typical legal systems and human behaviour. In fact, both the assumptions about citizens' willingness to comply in the absence of the use of enforcement mechanisms 57 and the assumption that a sense of group solidarity has increased after the terrorist attack are empirically grounded. 58 This already bars the complaints about unreliability voiced against the society of angels thought experiment. Sure, the scenario depicted is unlikely to take place. But an argument against the necessity of coercion need not appeal to what is likely; what is possible is enough. And given the proximity of this scenario to what we know about typical legal systems and human behaviour, all suggests that 56 We should assume that the system in place can still create and alter citizens ' legal positions (i.e. their rights, liberties, powers, duties, etc) and their legal status. Nothing in the scenario would prevent judges from (validly) declaring that someone's property has been transferred, or that someone is liable to pay damages, or even that someone has committed a legal wrong. However, despite having legal effect, there is a chance that these declarations are defied by some citizens given the lack of enforcement mechanisms. 57 Empirical work corroborates the idea that most legal mandates do not by themselves motivate citizens to comply with the law. Most citizens comply with most legal mandates most of the time in virtue of legal mandates matching their moral and cultural beliefs, and in virtue of a willingness to conform with social expectations. For helpful reviews and discussion of the empirical literature, see, e.g., the system at hand is a legal system, and not some other kind of institution. It seems, therefore, that even human legal systems are not necessarily coercive. A fortiori, legal systems are not necessarily coercive.

B. The Argument from the Fair Criminal System
Perhaps the biggest obstacle for accepting that a non-coercive human legal system is possible -a reason that may lead some to find the argument presented in the previous subsection implausible -is related to the important and salient role that punishment plays in our societies. Without punishing those who break the law, a large-scale institution such as a typical legal system is unlikely to persist. That is not because most people wouldn't be motivated to comply with legal mandates in the absence of punishment; they would. 59 The problem is that, as empirical studies suggest, those who are willing to cooperate and comply with institutional rules tend to withhold cooperation over time in the event they consistently see those who violate accepted institutional rules not being censured in some meaningful way. 60 It seems, therefore, that some forms of punishment must exist and must be applied to maintain a legal system working in a human society in the long run.
Of course, we can dispute two points at this stage. The first is that the considerations above only show that a legal system wouldn't persist for long if punishment is not administered. But nothing is said to account for the time in between the suspension of punishment and the disappearance of the legal system. If a legal system, however decadent it may be, is still present in this intermediate period, then we have all we need to conclude that a legal system that doesn't punish or provide for punishment is possible.
The second point that could be disputed is that the punishment must be carried out by legal institutions. In principle it seems possible (though perhaps undesirable) that, instead of providing for punishment and punishing, a legal system could just help determine who 59 See references on n 57. 60 See, e.g., Dominique JF de Quervain and others, ''The Neural Basis of Altruistic Punishment'', Science 305 (5688) (2004) has committed wrongdoing (respecting standards of proof and procedural guarantees) and let citizens or private organisations deal with wrongdoers. For the sake of argument, let's bracket these objections and concede both that a legal system would cease to be a legal system from the very moment punishment has been abolished or suspended, and that an institutionalised form of punishmentlegal punishment, as it were -is necessary for legal systems to exist in societies like ours.
Despite the importance of punishment to human societies, it would be a stretch to conclude that typical legal systems are necessarily coercive from the fact that they necessarily require punishment mechanisms to properly work. Punishment, as argued time and again by philosophers of criminal law, need not and should not be associated with cruelty. We should say the same about associating punishment with coercion: a legal system that provides for punishment and punishes need not (and should not) be coercive.
It all hangs on the way a legal system addresses its citizens when providing for punishment and when punishing those who break its laws. 61 A salient, and essential, feature of coercive actions is that the coercer does not address the coercee with respect. Coercers do not guide or attempt to guide the coercee's actions; coercers goad them. 62 Take the paradigmatic example of a coercive action: a scenario where a gunman demands a passer-by's purse. The gunman, in this scenario, does not warn or merely inform the passer-by of an unwelcome consequence. Instead, the gunman threatens to bring about an unwelcome consequence; he addresses the passer-by as merely a means to his own unilateral intention to get hold of the purse. It is in this sense that coercive actions are disrespectful: they 61 Of course, the effects entailed by the provision of punishment also determine whether punishment is coercive or not. As stated in the previous section (section IV.A), acts are truly coercive only if they generate specific effects: compelling, leaving people with no reasonable choice, etc. Determining the precise effects that an act must generate in order to be coercive is the job of a full account of coercion, which I am not advancing in this paper. 62 Stephen L Darwall, The Second-Person Standpoint: Morality, Respect, and Accountability (Cambridge MA: Harvard University Press, 2006), p. 49. are actions that characteristically deny or usurp the coercee's authority, power, or standing to decide to decline pursuing actions which the coercee did not choose or did not have reason to choose for herself. 63 Believing that our criminal law and our legal systems must address citizens in this peculiar way to be what they are would be mistaken. There is no reason to think that our legal systems must address citizens with the lack of respect typically associated with coercive proposals in order to fulfil the proper functions of punishment and criminal law. A legal system could communicate the wrongfulness of some actions, dissuade citizens from engaging in those actions, and censure those who engage in such actions without resorting to the kind of disrespectful treatment characteristic of coercive actions. All these functions could be fulfilled by a non-coercive criminal law system, without stopping the punishment of citizens. 64 To be clear, the legal (and criminal law) system I have in mind here isn't perfectly fair. The kind of system I have in mind is a system that addresses citizens with respect, and not merely as a means to the system's unilateral ends. It is the kind of system where the authorities are committed to implementing, upholding, and communicating a normative framework that citizens would have reasons to identify themselves with; a system that is committed to treating citizens as rational agents and making them aware of the consequences of their own choices -consequences citizens themselves would regard (or would have reason to regard) as adequate to their behaviour. It is a system where wrongdoers would be called 'to 63 Note that it is this characteristic lack of respect that explains why coercion is (at least) pro tanto wrongful and calls for justification. See, e.g., Arthur Ripstein, Force and Freedom: Kant's Legal and Political Philosophy (Cambridge MA: Harvard University Press, 2009), pp. 43-44; Japa Pallikkathayil, ''The Possibility of Choice: Three Accounts of the Problem with Coercion'', Philosophers' Imprint 11(16) (2011): 1-20; Darwall (n 62) 49-52. 64 In fact, while legal systems' conditional announcements that individuals will be punished (if they perform or abstain from performing some actions) can certainly be said to be coercive, it is not entirely clear if the act of punishing is the kind of act that can be coercive in and of itself. Perhaps the only way in which the act of punishing can be said to be coercive is indirectly. For example, when punishing someone sends a threat to others and compels them to adopt the coercer's preferred course of action to avoid punishment. But this raises many questions. First it is unclear if legal punishment is characteristically carried out with the intention to send this message and if it characteristically compels citizens to follow legal mandates. Second, it is also conceptually unclear if one can be coerced when one is not the addressee of the coercers' demand. What these remarks show is that the idea that punishment is closely associated with coercion should not be taken for granted and require argument. answer to their fellow citizens' 65 for wrongs that wrongdoers themselves accept (or have reason to accept) as wrongs. 66 We may end up never completely getting rid of coercion and successfully implementing such a moderately fair legal system. But this doesn't mean that doing so is impossible -metaphysically, conceptually, or even humanly. And if we happen to implement this non-coercive and reasonably fair system, it would still be sensible to regard it as a legal system. In fact, a non-coercive legal system of this kind would not just be a legal system; it is perhaps the kind of legal system humans -not angels -ought to implement.

V. CONCLUSION
The traditional view according to which legal systems are not necessarily coercive is closely associated with the society of angels thought experiment. This association has led some to question this view's plausibility and end up rejecting it altogether. I argued in this paper that the traditional view is sound. First, none of the objections against the reliability of the society of angels thought experiment succeed. Second, there are (at least) two arguments that go further than the traditional view and favour an even stronger view according to which even typical legal systems -legal systems made by humans to address human needs -are not necessarily coercive.
These arguments, presented in section four, rest on relatively thin assumptions: empirical facts about human behaviour and on widely accepted assumptions about coercion and coerciveness of legal systems. If successful, they show that most who have assumed that typical legal systems are necessarily coercive do not seriously take into account what legal systems can achieve without resorting to coercion. Typical legal systems already owe much of their effectiveness and stability to factors other than the threat of sanctions and enforcement mechanisms. 67 But typical legal systems can do better and must do better. Discarding the idea that coercion is needed in our societies is a good starting point towards a less disrespectful and less wrongful form of legal governance.
It is unclear if my arguments will convince critics of the traditional view. But the arguments presented in this paper might at least help in moving the discussion away from angelic scenarios; something critics seem to have longed for.