Behavioural Evidence,Yes;Normative Behaviourism,No

Abstract

Floyd defends normative behaviourism against ‘mentalism’. His characterization of political philosophy as mentalism is uncharitable, and it is not clear that normative behaviourism provides greater evidence of convergence that we find in liberal political philosophy. To interpret behaviour, one must theorize the effects of institutions on that behaviour, it is therefore problematic to defend institutions on behavioural grounds alone without ‘mentalistic’ theory. Normative behaviourism uses a ‘contingent imperative’; however, this leaves the behaviour normatively undefended. A potential response by Floyd to these criticisms depends upon misinterpreting Cohen’s argument that fact-free principles underlie all policy recommendations. Floyd’s own recommendations require at least one fact-free principle in Cohen’s sense. Floyd is correct that behavioural evidence is important to political philosophy.

Jonathan Floyd’s (2017: 1) book makes the claim that we should not try to derive political principles from our ‘pre-existing thoughts’, since they are too messy to be systematized into political principles. He calls this ‘mentalism’. The problem he sees with current political philosophy is that there is disagreement over ‘how we should live’ (Floyd, 2017: 26). He says ‘political philosophers should pay more attention to the behaviour of real citizens than to the reflections of other political philosophers when assessing the quality of different political systems’ (Floyd, 2017: 1). He calls paying such attention ‘normative behaviourism’. ¹ If behaviour is more consistent than philosophers’ positions, then we might have a better chance of agreement.

I first query whether there is more agreement in action than thought, and then argue that, even if so, we cannot derive normative political principles from behaviour alone. That constitutes my claim ‘normative behaviourism, no’. I will then argue that political philosophy should utilize evidence about political behaviour, defending a version of Floyd’s account that is weaker but perfectly consistent with standard ‘mentalism’ − though not, perhaps, with ideal theory.

Normative Behaviourism, No

Floyd asserts that there is more consistency in behaviour than in political argument in at least one restricted realm. He claims that the fact there is less sedition in liberal democracies than in autocracies shows people prefer liberal democracies. Is it true there is less sedition in democracies? Some of what counts as sedition in autocracies is simply regarded as part and parcel of liberal democracy. In democracies, you are allowed to publish anti-government tracts, and engage in peaceful street protest. If one were to simply tally the numbers of such anti-government actions, then almost certainly there would be more in liberal democracies. ² Or think of the behaviour of Trump and Bolsonaro supporters following an election loss: an attack on the very principles of liberal democracy. However, because it costs more to revolt under autocracy than in a liberal democracy, we cannot easily compare these actions in terms of what they show us about preferences for regimes.

This is the first lesson of revealed preference analysis. Behaviour needs interpreting in terms of the environment in which decisions are made. With enough variation in choice and environments, we can construct plausible utility functions for types of people, but comparisons across institutional forms are difficult. Floyd’s other example is of crime being lower in more egalitarian societies, suggesting people prefer more egalitarian societies. Well, certainly, we should expect the poor to. However, consider Tsebelis’ (1990, 1991) game-theoretic argument, which suggest an equilibrium crime rate. In his model, if the penalties for crime go up, the rate of criminal activity goes down; but as it goes down, society spends less money on crime intervention, making criminal activity easier, so that, the crime rate goes up. Equilibrium occurs at the rate the community finds acceptable. Less egalitarian societies might have higher crime rates because voters (who tend to be richer) can spend private money protecting and insuring themselves. It is the acceptable equilibrium rate that differs across societies. The lesson is that, given preventing crime costs money, the crime rate reflects community demand.

Despite these doubts, I will accept Floyd’s claim that more people in authoritarian societies are unhappy with the regime than in liberal democracies. Does this show more consistency over these issues than among mentalist political philosophers? Surely, there is near-unanimity among political philosophers, at least those who write in the circles that Floyd and I inhabit, that liberal democracy is preferable to autocracy. There is more variation over the merits of freedom and welfare among philosophers, but is this higher than we see in political behaviour? Voting behaviour does not suggest that people want more redistributive policies in most liberal democracies. Thus, I simply do not accept that examining behaviour gets us closer to agreement than mentalism.

In any case, we cannot derive normative principles from behaviour alone. Floyd’s argument for this claim trades on his uncharitable characterization of political philosophy as drawing political principles from ‘pre-existing thoughts’. I see political philosophy as composed of arguments justifying conclusions, even if those arguments defend or modify moral intuitions.

The logic of normative behaviourism consists in what Floyd calls a ‘contingent imperative’ that parallels the logical form of a hypothetical imperative. He says, ‘Consider Kant’s distinction between “categorical” and ‘hypothetical” imperatives. Clearly the kind of imperative discussed here is essentially the latter type, which claims that if you want x, then you should do y’ (Floyd, 2017: 176). Now, of course, Kant did not think morality was composed of hypothetical imperatives, only categorical ones; and one reason is that a hypothetical imperative justifies its conclusion (the clause after ‘then’) and not its hypothesis (the clause between ‘if’ and ‘then’). So, the hypothetical imperative justifies y and not x. If the hypothetical imperative is a moral and not simply a practical one, then we need to defend ‘wanting x’ as a moral objective.

Floyd’s (2017) contingent imperative states: ‘given you have a set of preferences x, you should adopt or support or accept set of principles y’ (p. 176). The preferences (x) are inferred from behaviour (sedition and committing crimes) to the conclusion that we should adopt the principles of liberal democracy and equality. Yet are these preferences moral ones? If I am rich and vote for low taxes, is my desire for low taxes a moral one? If the majority of people votes for low taxes, does that mean the desire for low taxes is a moral one? Floyd’s account of generating normative principles from his contingent imperative only plausibly parallels mentalism because he assumes mentalism generates normative principles from undefended ‘pre-existing thoughts’. But it does not. Where hypothetical imperatives are used, the hypotheses are defended with other arguments.

Behavioural Evidence, Yes

Social science gave behaviouralism a bad press in the 1960s and 1970s, for being untheorized. These days it is recognized that how we interpret behaviour requires theory (Dowding, 2019). I might vote for low taxes because I earn a lot, or because I fear that high taxes will have a disincentive effect on effort and will result in lower growth and I think we need higher levels of growth. If I do not turn up to help drain the field, that might be because I do not want the field drained, or it might be that I do want the field drained but think enough others can do it without me, or that turning up is a waste of time because not enough others will turn up. We cannot assume from lack of action on climate change that people do not care about climate change. We need to collect data and theorize carefully about behaviour to work out preferences.

Even if we accept that Floyd’s interpretation of sedition and crime rates demonstrates preferences for liberal democracy and equality, respectively, that interpretation involves a mentalistic exercise, interpreting what is going on in the head of actors. Floyd (2017: 175) associated normative reasons with ‘internal’ ones (Williams, 1981). ³ Does that not make him a mentalist after all? The answer depends in part on whether Floyd wants to distinguish internal moral from non-moral reasons. If we include all reasons for action as part of the moral calculus, then we are representing preferences as von Neumann−Morgenstern utility functions, and Floyd becomes some sort of utilitarian. That commits him to accepting that all behaviour-based preference is equal in the moral calculus. But is that Floyd’s position? He seems to assume sedition and crime are wrong. Does he have a ‘pre-existing thought’ about this? His claim is that people do not want to commit crimes or seditious acts because they are risky. That is why they are evidence that there is something wrong with social institutions. However, von Neumann−Morgenstern utility functions are constructed through assumptions about expected gains given the risks of any action. Again, this makes comparison across different societies in terms of ‘human’ preferences problematic. Our preferences are, in part, formed by the institutional structure we inhabit.

Rather, for ‘do x’ derived from ‘if you want y, do x’, to be a moral imperative, y has to be morally desirable, not simply something someone wants. Furthermore, the nature of the moral desirability of y will affect what ‘xs’ are justified in attaining y. Reasons for y, not simply their priority under some choice conditions, matter. For example, say constant electronic surveillance of all behaviour reduces crime and sedition even more than social democracy. Furthermore, suppose that fact holds to the same degree no matter what the form of social organization. We cannot conclude we want more surveillance just because that leads to less crime and sedition, because we also have preferences over surveillance. Behaviour and institutions interact, and we cannot generate institution-free conclusions without other, normative assumptions. Missing from Floyd’s defence of social democracy based on the priority of choice are reasons; providing such reasons is why, in Floyd’s terms, we are mentalists.

Behaviour and institutions interact, and we cannot generate institution-free conclusions without other, normative assumptions. Institution-free assumptions are often used in political theorizing − the chooser has ‘complete information’, ‘lacks knowledge of their own preferences’, ‘shows perfect empathy’, and so on. ⁴ Crucially, however, such models can also prove that certain sets of desirable features cannot be simultaneously satisfied. Arrow (1951) demonstrates the incompatibility of desirable features of decision mechanisms; Sen (1970) of aspects of welfare and rights; Bowles (2016) that efficiency and voluntariness are only compatible if we restrict allowable preferences. Behaviour here cannot demonstrate priority, since institutions will always favour behaviour using one or other of the desirable features.

I believe we can separate two aspects of political theorizing, though each feeds into the other. The first concerns the moral content of politics. It involves metaethical debates − for example, the relative importance of liberty versus welfare; or conceptual issues, what is liberty, or what sorts of considerations go into human welfare. Given metaethical disagreement and the fact of plural values, ⁵ the second aspect of political theorizing concerns ‘what institutions are required for us to live peacefully together?’.

These are separate aspects, since a libertarian and a welfarist could agree on the second institutional question, despite fundamental disagreement over the first. For example, they could both agree that we need governing institutions that are democratic, with relatively politically insulated legal systems, a free press, and various citizen rights and liberties. They feed into each other, since they might disagree about what rights and liberties citizens should have, what a free press entails, what should be politically insulated by law and what form democracy should take. Some libertarians, for example, think many issues should be subject to individual veto and not majority vote (Buchanan and Tullock, 1962), surely something no welfarist could agree to. So, while the two aspects can be separated, moral disagreement impinges on considerations of the second. They cannot be forced together. ⁶

Floyd thinks political philosophy (mentalism) fails because it does not reach agreement on the metaethical or plural value issues. No moral pluralist thinks this is a problem. I do not see it as a failure, since I take the arguments on each side to be part of political debate about how we want society governed. They provide motivations for the very behaviour that Floyd wants to use as evidence for what we ought to want. Where Floyd is correct is that where we all largely agree on the outcome, the fact we have different reasons might not matter. ⁷ What we often find is that different reasons justifying the same evaluation come to matter when we apply those reasons to a moral dilemma over which we disagree. For value pluralists, moral dilemmas just show that there are competing moral values; that is a fact of human existence (Berlin, 2002; Lukes, 1991; Williams, 1981).

Claiming political argument is otiose is to say we should not consider morally motivating reasons for action, nor embark upon highfalutin political dispute. For this aspect of political philosophy is the high end of democratic political argument. Now Floyd never actually suggests we should not engage in political argument. After all, he is a democrat. But he does suggest political philosophy fails because there is no hope of agreement. However, the post-Rawlsian project has been not about reaching agreement over moral matters, but providing a constitutional (peaceful) way of managing conflict − how we govern ourselves. As I suggest above, I think there is at least as much agreement about that among mentalist philosophers as we can discern from public behaviour.

For normative issues themselves, we can see, within a constitutional framework, it is a political game and, naively, we suppose that those with the best arguments will, at least temporarily, win. Less naively, those with the best arguments will help society move towards the vision they promote. ⁸ Importantly, however, it is a game, and games need rules. If we cannot agree on the rules, then we cannot have political argument (Dowding, 2013). We can only have war. And this is serious. It is attacks on the rules as they have been understood that is today leading the US towards civil war.

The second aspect of political philosophy is about the rules of the game. Here we need agreement, or at least compromise, if we are to be able to engage in political argument. We need agreement on the second aspect, to peaceably disagree when undertaking the first. Floyd’s account is exclusively about this second aspect. His book is a defence of social democracy. The problem, however, is that he wants to justify social democracy without any political argument. In his account, we just have to take low crime and no sedition as evidence of satisfaction; but, as I have argued, that cannot be right. Some ways of attaining desirable ends are not acceptable, shown by my (mentalist) thought experiment.

Floyd demurs, on the grounds that any public justification is a part of behaviour and thus a fact. This is a curious argument which I cannot discuss at length here. However, Floyd seriously misinterprets the claim of his main target (Cohen, 2008) that all moral principles are either fact-free or ultimately depend upon some underlying fact-free moral principle. Floyd (2017:171) claims that principles we hold are mental and thus depend upon ‘facts about the thoughts in our head’. Cohen does not deny that normative principles are factual in that sense or that truth-apt statements, including true principles, represent facts. The distinction Cohen (2008: 229) makes is:

A normative principle, here, is a general directive that tells agents what (they ought, or ought not) to do, and a fact is, or corresponds to, any truth, other than (if principles are truths) a principle.

Whether those propositions have ontological form in a particular mind at any time is irrelevant. ⁹ Importantly, Cohen thinks we can rationally defend such principles (they are not simply ‘pre-existing thoughts’). Moral dilemmas occur when we have to trade such normative principles. Much of political philosophy is about trying to provide arguments about how such trades should, in general, be made.

I think Floyd is correct that political philosophy should take more account of empirical generalizations about human behaviour. But I think we need to do so in a manner that is not necessarily appropriate for moral argument. Moral philosophy needs to ignore certain inconvenient facts about ourselves. When we construct arguments about what, morally, we ought to do as individuals, we need to ignore the fact that we are self-interested, weak-willed and sometimes downright immoral. The fact that I know that I am weak-willed, and therefore will probably not do what I ought to do, is not relevant to what I ought to do. Indeed, I take it that the fact I am weak-willed, and so will probably not do what I ought to do, is precisely why my consideration of what I ought to do should ignore that fact. Morality is about giving us reasons to do that which we know, without those reasons, we would not do. It cannot therefore simply track behaviour. And it gives those reasons even if we know we still might not do what those reasons suggest. Embarrassing people into collective acts through moral persuasion is sometimes necessary to motivate them.

Political philosophy, however, is different. Designing political institutions requires us to consider inconvenient facts about human behaviour. That is true, even when every single person agrees what every single person ought to do under all conceivable circumstances. We need to think about the incentive effects of tax rates, for example, or how food regulations affect food choice and thus obesity rates in society (Dowding, 2020). One of the defining features of politics, however, is that people have different preferences, both self-regarding − what they want for themselves − and other-regarding − what they want for the good of all. Designing political institutions requires recognizing both the fact (and degree) of backsliding and the fact (and degree) of plural values. For the first, we witness patterns of human behaviour under different incentive structures; for the second, the degree of cultural and value heterogeneity in any given polity. Our design needs to recognize not only how people behave, but also how we think their values should be represented.

I believe too much political philosophy ignores my distinction between moral and political considerations. So, I applaud Floyd’s attempt to bring behaviour into political theorizing. Political institutions motivate and police social patterns of behaviour. The degree of moralism or political realism in those outcomes depends upon balancing moral considerations with those of political stability, legitimacy and exigency. I take it, therefore, that our moral arguments are relevant to what we want those outcomes to be. Hence, ‘Behavioural Evidence, Yes; Normative Behaviourism, No’.