For the Love of Spinoza

Happy New Year everyone. Levi recently put up an interesting post about Spinoza’s account of the relation between causal knowledge and ethics (here). As some of you may know, I’m quite a big fan of Spinoza. Not just of his metaphysics, but also of his resistance to Aristotelian teleology and his resolve to think freedom in a way compatible with his completely deterministic metaphysics. As I’ve argued elsewhere (here), Spinoza reconciles freedom with the principle of sufficient reason in a much healthier manner than Leibniz, and a lot of contemporary debates on this issue can be interpreted as taking place between neo-Leibnizians and neo-Spinozists. I’m firmly in the neo-Spinozist camp, but this doesn’t mean that I agree with Spinoza completely. Levi’s post very clearly outlines one of the points where I have an important disagreement with him (and his heirs), so it’s useful to address it. It also gives me a good excuse to work through some of the ideas I’ve been having about ethics and politics over the past few months.

This post is another fairly long one (8,000 words or so), but it not only contains my thoughts on Spinoza, but also some thoughts on Kant, Foucault, Sellars, Hegel, and Plato, which it pulls together to provide the outline of a theory of Justice. That may sound a bit over the top, but I’m nothing if not ambitious. Anyway, on with the show…

1. A Spinozan Problem

The central thesis that Levi examines is the idea that as our understanding of nature increases, we can gradually replace the ethical prescriptions (or norms) we inherit from the social traditions we find ourselves in with causal claims about the effects that the prescribed (or prohibited) actions have. To quote Levi:-

Spinoza’s idea seems to be that the more causal knowledge we have about our bodies, our psychology, world, and how the social world functions the more we’ll be able to dispense with ethical commandments. Here it’s important to be clear. Spinoza’s thesis is not that the more causal knowledge of body, mind, world, and society we develop the more unethical we’ll become. Rather, Spinoza’s claim is that as we acquire knowledge of body, mind, world, and society our reasons for doing things will change.

The idea here is that improved causal knowledge enables us to explicate the reasons why we follow these norms, assess these reasons, and then potentially revise our norms in accordance with this assessment. On the face of it, this kind of dynamic ethics is a very good idea, and Levi contrasts it to what he takes to be a more static deontological approach:-

The advantage of norms or commandments is that they are able to motivate behavior in the absence of the person possessing causal knowledge. The person who lacks knowledge of shellfish and how it’s likely to interact with the body will, if they accept the authority of the person or deity issuing the commandment, avoid eating shellfish. The problem with deontological approaches to ethics is that in their structure as commandment they tend to foreclose any way of evaluating commandments to determine whether they are well founded. The prohibition against eating shellfish makes this point clear. If my reason for not eating shellfish lies in God’s command, then anything I might learn about the properties of how my body and shellfish interact is irrelevant to whether I ought to eat shellfish. The prohibition is absolute and there are no circumstances under which I should eat shellfish. However, if it turns out that moral commandments are really dimly perceived causal claims, then it follows that further knowledge of how my body and shellfish interact, coupled with the development of safe ways of preserving shellfish and of evaluating whether they’re safe to eat could lead the community that formulated this prohibition to abandon it. Something along these lines seems to have taken place in contemporary society surrounding prohibitions against premarital sex, sex outside of marriage, and sex for the sake of pleasure. In a society where there’s no reliable birth control, these prohibitions make good causal sense. However, with the development of reliable forms of birth control, they no longer make sense.

I think this sets up a false dichotomy between norms as commandments whose force resides in some special authority, and norms as shorthand forms of causal knowledge that can ultimately be eliminated in favour of detailed and explicit causal analyses of ourselves and our environment. Not only do I think it is possible to occupy a middle ground here, I also think it is necessary to do so, because neither alternative is tenable.

As I think Levi would agree, Socrates’ challenge in the Euthyphro holds good. What is pious/moral/good must be so because it is, not because it is said to be by anyone. This is what defines the specific normative force of ethical demands (e.g., the prohibition of murder), as opposed to the normative force of various social conventions (e.g., dinner party etiquette). This is also to some extent commensurate with Kant’s conception of autonomy: not even God himself should be able to bind me to a norm unless I consent to be so bound (in some fashion). Ethical norms are then those norms that we cannot but help be bound by, insofar as they are derived from the universal structure of binding itself (again, in some fashion). This should indicate that, at the very least, deontological approaches derived from Kant’s account of autonomy don’t interpret ethical norms as commandments in the way Levi takes them too. However, regardless of whether one accepts this transition from Socrates to Kant, we can all agree that the model of ethical norms as commandments is untenable.

Where Levi and I (possibly) disagree is on the possibility of eliminating talk of specific ethical norms in favour of causal knowledge. To see why, it’s necessary to further unpack Spinoza’s account of the basis of normativity. The core feature of this account is his attempt to replace the absolute distinction between good and evil with the relative distinction between what is good and bad for a given mode. Although this is often read as the wholesale annihilation of teleology, there’s an important sense in which it is actually the transition to an immanent teleology. Whereas Aristotle attached different purposes to things/modes on the basis of the types of things/modes they are (e.g., each human’s telos qua human is to flourish), in such a way that there must be a transcendent assignment of purposes by a privileged unmoved mover, Spinoza attaches the same purpose to each thing/mode qua thing/mode, namely, to maintain itself in accordance with its essence. This translates into a dual imperative to survive and to maximise one’s power (understood as capacity). Whether something is good or bad is then judged relative to the perspective of a particular mode in satisfying this dual imperative: shellfish may be bad for me (because I’m allergic to it), but not necessarily for others, and high taxes may be bad for me (because I’m an ‘entrepreneur’), but good for society considered as a whole. This means that all normative force is derived from a simple and ontologically egalitarian principle.

It’s important to see how this derivation works though. The reason it underwrites the elimination of independent ethical principles in favour of causal principles is that it essentially transforms all ethical reasoning into instrumental reasoning. This is to say that all ethical reasoning becomes a matter of determining appropriate (and inappropriate) means to achieve the ultimate end of maintaining oneself in accord with one’s essence. It’s the fact that instrumental reasoning deploys facts about the causal factors involved in achieving the end in question that underwrites the elimination of distinct norms (I’ve written more about how such practical reasoning works here). Prima facie, there are two dangers that this approach must avoid:-

i) It must avoid collapsing into a form of egoism, wherein everyone is justified in doing anything to survive and maximise their power.

ii) It must avoid suppressing the individual’s freedom to determine their own rational ends, by preventing individuals from choosing to subordinate their own lives (and/or their own empowerment) to the service of something greater.

Spinoza avoids both of these dangers, but only up to a point. He does this by recognising that human beings compose larger modes, such as cities and states, which operate at a higher scale. He uses this to curb the potentially egoistic implications of the individual’s imperative to preserve itself, by subordinating it to the imperative to preserve the larger mode of which they are a part. Although it is still important to do what is good and avoid what is bad for me, it is more important do what is good and avoid what is bad for the state. This might sound somewhat authoritarian, but it is not insofar as the freedom of the individual is not supposed to be completely dissolved in that of the state considered as an individual entity in its own right. It’s simply that the norms governing the individuals who are part of the state are supposed to be justified in accordance with the role they play in maintaining the state in accordance with its essence, and this essence need not be authoritarian in form. The norms are thus in some sense reducible to causal facts about the social structures that constitute the state as a mode in its own right. However, as I said, this only avoids the above dangers up to a point.

The first issue is one of choosing the correct perspective from which to judge what is good and bad. There are a whole host of different modes to choose from here: there’s what’s good for me qua biological organism,what’s good for me qua person whose extended presence outlives this organism (be it in the form of a mind uploaded into software, or the records and memory-traces we leave behind), what’s good for my familial line (i.e., as a loose alliance of selfish genes), what’s good for any one of the socio-cultural groups I belong to (e.g., class, intellectual school, corporation, etc.), what’s good for my society (e.g., city, region, state, etc.), what’s good for my race/species (e.g., caucasian, human, primate, etc.), what’s good for the biosphere I belong to, and so on all the way up. Although we can possibly see some of these as being nested, and thus as offering neat relations of subordination (e.g., organism < familial line < species, or person < class < society), there are plenty of conflicts that can’t easily be resolved simply by invoking scale (e.g., organism vs. person, or region vs. class). There’s also no natural ceiling at which subordination stops. There are many who would subordinate the survival of our species to the survival of our biosphere, but there are very few who would defend a similar subordination of the survival of our biosphere to the survival of our solar system, yet Spinoza’s minimalist teleology provides no good way to block this result. There is something really intuitive about Spinoza’s proposal, but it simply lacks the resources to adequately individuate which modes we’re determining the good in relation to.

The second issue is related to the first, but it combines the problem of perspective with the issue of rational self-determination (danger (ii) above). By articulating the subordination of what is good for one mode to another in mereological terms (i.e., the needs of the parts are subordinated to the needs of the whole), Spinoza does allow us to subordinate our own lives to something greater, but it’s ontological foundation excludes the role of free choice in determining the collectives we wish to be part of. It should be obvious that I don’t mean some weird ontologically circumscribed free will here. I’m not complaining about Spinoza’s determinism, as this is precisely what I think is most laudable about his thought. The point is that which higher scale modes I count as part of (and thus should subordinate my own survival/empowerment to) is entirely a matter of how causally enmeshed I am in them, and has nothing to do with which groups I choose to identify with. This is reminiscent of some rather crass arguments regularly thrown at anti-capitalists: you cannot morally disentangle yourself from capitalism until you have causally disentangled yourself from it (i.e., go live in a hut in the middle of nowhere, stop drinking coffee, and forget about using the internet, you lousy hippies…). Although causal facts about the way we are enmeshed within various socio-economic structures should certainly play a role in determining our responsibilities to and on behalf of the collectives we’re part of, they can’t be made the only factors without abandoning the crucial normative role of self-determination in justifying political change. We must have at least some autonomy in determining which responsibilities we are bound by.

I think this shows that, although Spinoza’s approach has some promise, it simply isn’t tenable as it stands. It’s important to point out here that this situation isn’t improved by Deleuze and Guattari’s transformation of Spinoza’s ethical project. What they do is to essentially eliminate the role that individual essences play in Spinoza’s account. Instead of surviving by maintaining (and empowering) ourselves within the fixed limits determined by my essence, they enjoin us to shed the limitations that our structure imposes upon us (deterritorialisation) not with the aim of dissolving ourselves into the whole, but with the aim of restructuring ourselves so as to be more adaptable. This is a kind of empowerment unbounded by the limits of any putatively pre-determined essence. However, simply changing the specific nature of the end that is immanently bound up in each entity does not do anything to overcome the problems posed above. Spinozism on its own is not enough. If it is to live up to its promise, it must be supplemented.

2. A Kantian Supplement

It’s no secret what I think this supplement is. I’ve explained my belief that the structure of rational agency forms the basis of ethics  in several places (see here, here and here). I’ve also explained my belief that this means granting ethics (and the normative as such) complete independence from metaphysics (see here and here). In both these respects I’m resolutely Kantian. However, I don’t endorse the specifics of Kant’s own ethical philosophy (I reject the categorical imperative as he formulates it). Although I’ve said the odd positive thing about my views on ethics and politics (see here, here and here), I’ve yet to present anything like a complete alternative to Kant’s own deontological approach. I won’t attempt to do so here, but I will take this opportunity to pull some of the ideas I’ve discussed before into a more coherent outline. The problems with Spinoza’s approach discussed above provide the perfect context for addressing some of the ideas I’ve been developing on this front.

The first thing to tackle is the transition from ontology to deontology. As we explained above, Spinoza’s reduction of the normative to the causal is based upon an immanent telos he locates in every mode qua mode. This makes everything, from my cells, my organs, my body, my family, my city, my country, to the earth as a whole into an ethical subject of some sort. The first move we make is to deny this, and to accept that only some things are ethical subjects. However, the distinction between things which are and things which aren’t ethical subjects is not an ontological distinction of any kind. It is not a matter of having a special mode of Being (e.g., Existenz), but rather about occupying a certain sort of normative status, which means playing a certain role within a norm governed social practice. This is being counted as a rational agent, which is to say, as something that can be treated as responsible for its beliefs and actions, insofar as it is able to provide reasons for at least some of them.

This means that being a rational agent (and thus an ethical subject) is less like belonging to the species homo sapiens, and more like qualifying as a legal adult in a particular society. It might be that in both cases the circumstances under which someone is so classified are completely objectively assessable (e.g., having a certain genetic characteristics, and being over 18 years of age, respectively), but this is not the case with the consequences classifying someone as a legal adult, which prescribe certain ways of acting on their part, and on the part of others in relation to them. It is not anything about the intrinsic features of 18+ year olds which grants them this normative role within our practices, but rather a decision on behalf of that society (though it is of course possible to tell a causal story about how these features were involved in the production of that decision).

However, there are some important differences between rational agency and legal adulthood. First, although there’s a good sense in which the status of legal adulthood is constitutive for the whole practice of instituting laws, the status of rational agency is constitutive for the practice of instituting any norms at all. Whereas the role of legal adults is something that we’ve instituted, the role of rational agents is a transcendental condition of instituting any roles whatsoever, and as such is not itself instituted. This means that the consequences of classifying someone as a rational agent aren’t exactly decided by us, even if they’re not exactly objectively assessable either. Second, the circumstances under which it is appropriate to count something as a rational agent are a lot more complicated. We cannot simply choose to classify the chair I am sitting on as a rational agent, and ascribe to it various rights and responsibilities on this basis. We can only classify things as rational agents which have certain specific capacities, namely, the capacities to track theoretical and practical commitments and the inferential connections between them necessary to play the game of giving and asking for reasons. This has something to do with the idea that ought implies can, but I won’t go into that debate here.

The difficulty here is that these capacities must be specifiable in abstract functional terms, much like the abstract description of computational capacities provided by Turing machines, recursive functions and lambda calculus. This is because they are essentially capacities for manipulating tokens (i.e., sentences) the material constitution of which is completely irrelevant. Just as one can instantiate a computer in pressed silicon wafers, intricate networks of neurons, or elaborate systems of cats, mice, gates, pullies, and bits of cheese, you can instantiate a rational agent in anything that can be assembled so as to display the correct functional relationships. The million dollar question is exactly what this functional description is. I’m not going to provide a complete account of it here, as I haven’t got one. This shouldn’t be too surprising, given that to do so would essentially be to solve the problem of true AI outright (though another way of describing it would be completing Kant’s project of transcendental psychology). Nonetheless, there are a few things we can say about what this description would look like.

A rational agent is abstractly characterisable as a causal system that deals with two kinds of inputs and outputs: sentential inputs/outputs (S) and non-sentential inputs/outputs (NS). We can thus decompose it into subsystems, which fall into four distinct types:-

i) Reasoning Systems: these are subsystems that take sentential inputs and produce sentential outputs (S -> S). These are the mechanisms that are involved in keeping track of our theoretical and practical commitments (or beliefs and intentions, if you prefer), and updating these commitments on the basis of the inferential relations of consequence and incompatibility that hold between them. It’s theoretically possible to model these in a whole bunch of different ways, using the same resources deployed by computing theory to describe symbol processing systems in general. There’s way more technical detail to work out than I could reasonably go into here.

ii) Perceptual Systems: these are subsystems that take non-sentential inputs and produce sentential outputs (NS -> S). This is obviously a way of looking at the Sellarsian account of perception in terms of Reliable Differential Responsive Dispositions (RDRDs), though it’s also obvious that Sellars’ account requires that such subsystems (or dispositions) tie into reasoning subsystems in order to count as perception, rather than mere differential response. They translate events (stimuli) into sentences that denote the agent’s theoretical commitments.

iii) Actional Systems: these are subsystems that take sentential inputs and produce non-sentential outputs (S -> NS). Again, this is obviously a way of cashing out Sellars’ account of action in terms of RDRDs, which equally requires them to be tied into a reasoning subsystem of some sort. They translate the sentences that denote the agent’s practical commitments into events (actions).

iv) Coping Systems: these are subsystems that take non-sentential inputs and produce non-sentential outputs (NS -> NS). These might seem to be unrelated, insofar as they don’t involve sentential inputs or outputs, but they’re incredibly important. It’s understanding the role that these subsystems play that lets us reincorporate Heidegger’s insights about the practical basis of intentionality, insofar as Heidegger’s practices for dealing with equipment are just these kinds of coping system in which we move directly from sensory input (stimulus) to motor output (action). However, not all coping systems are Heideggerian practices so defined. On the one hand, only some processes involve the possibility of what Heidegger calls circumspective interpretation, which is essentially adaptive feedback (e.g., dynamically adjusting the force applied to a tool), and fewer still involve adaptive feedback that can proceed via a reasoning subsystem (e.g., reconsidering the force applied on the basis of a quick calculation), which Heidegger would call non-circumspective interpretation. There’s a lot more to say here, but I’ll cut it short.

Interestingly, all of these systems can be judged in terms of their reliability. Reasoning systems can be more or less reliable in computing outputs from inputs. This kind of reliability is what enables us to make sense of the notion of heuristics. Perceptual systems can be more or less reliable, and this is what underwrites an agent’s observational expertise. Actional systems can be more or less reliable, and this is what underwrites an agent’s practical expertise, although there is a blurry line here between the reliability of actional systems and the reliability of coping systems. This is related to (though not necessarily the same as) the distinction between conscious and unconscious practical expertise. All of these kinds of reliability can be understood to be forms of practical expertise in a broad sense, if we distinguish between broad and narrow senses of ‘practical’ by talking about causal systems in general (with unspecified types of inputs and outputs) and those systems that specifically produce non-sentential outputs (actional and coping processes). This is how we cash out Brandom’s (and Heidegger’s) idea that theoretical understanding is based in practical understanding, because broadly theoretical understanding consists in the possession of abilities (reasoning systems and perceptual systems) that are instances of practical abilities in the broad sense.

For a system to be counted as rational, it must not only be capable of both taking and producing sentential and non-sentential inputs and outputs, but the ways in which it consumes and produces these inputs and outputs must be subject to assessments of all four kinds of reliability in appropriate ways. First and foremost, it must possess a Core Reasoning System (CRS), or a highest-level reasoning subsystem that is not itself a subsystem of another reasoning system. This reasoning system tracks its theoretical and practical commitments and entitlements, and corresponds to what Brandom (following Kant) would call the transcendental unity of apperception. The CRS must reliably track its commitments. This relates to some of what I’ve written elsewhere (here) about sandbox responsibility. We treat children (and some of the mentally incapacitated) as having a kind of privative rational status precisely insofar as they aren’t reliable in this respect.

The Sellarsian conditions that we place on top of this are that the system has a set of reliable perceptual and actional systems that are correctly connected to the CRS (the second condition is what differentiates us from Parrots with RDRDs to say words in response to perceptual inputs, and dogs with RDRDs to produce actional outputs in response to words). The Heideggerian conditions are then that we have reliable coping systems (practices for dealing with things in the world) that are connected up with the CRS in such a way that the practical understanding implicit in them can potentially be deployed in reasoning (both practical and theoretical). This is something I’ve discussed more elsewhere (in section 4 of this post). The issue of the connection of perceptual and coping systems to the CRS relates to Kant’s discussion of the difference between empirical and transcendental apperception, and the fact that for any representation to count as conscious there must be the possibility of us becoming self-conscious of it (i.e., of bringing it under the transcendental unity of apperception, or of it feeding through the CRS). This is the terrain on which Brandom and McDowell’s debate on the nature of perception, and the status of unendorsed yet conceptually articulated contents is to be fought (see Brandom’s most recent paper on the topic here).

I must reiterate that this is a very sketchy outline of a functional description of rational agency, but it should be enough to demonstrate how we should proceed in providing a more complete account. What is required is a sort of abstract diagram of the various subsystems (and the relations between them) necessary to provide the minimal capacities characterisic of a rational agent. This would give us the essential resources required to distinguish those things that count as ethical subjects from those things that do not.

3. A Hegelian Spirit

Now we’ve made the move from ontology to deontology, we can reconsider some of the Spinozistic themes discussed earlier. The first thing to ask is whether there is a deontological parallel to Spinoza’s ontologically egalitarian telos of survival and empowerment. I think there is. I’ve written a bit about my interpretation of Foucault’s ethics of self-construction before (here), and I think he presents us with this parallel, in the form of a strange blend of Kant, Nietzsche, and Greek virtue ethics. To quote myself:-

[Foucault’s ethics] is a Kantian aesthetics of existence in which we explicitly take up individuation as a structural end of practical reason itself, engaging in a continuous process of self-construction with the only purpose being that we become more able to construct ourselves, thereby intensifying the causal dimension of autonomy, or positive Freedom. It is essentially a far more developed version of both Heidegger’s ethics of authenticity, and Nietzsche’s ethics of adaptability. It is thus the natural development of virtue ethics, except that it moves beyond the notion virtue and focuses on the rational core of the notion of agent-centred ethics.

The core idea here is that reason itself demands that we experiment upon ourselves, as if constructing ourselves as a work of art, but always with the aim of maintaining and increasing our freedom to do so. This is to say that we should, by default, make our choices in such a way that we maintain/increase our capacity to choose, both by maintaing/increasing our capacities for action (and thus what options are available to us), but more fundamentally by maintaining/increasing our capacities for reasoning (and thus our ability to effectively decide between these options). The latter are more important insofar as they provide the causal conditions of all choice, not just the particular choices that new kinds of practical ability facilitate. The parallel with Spinoza here should be obvious. However, instead of being obliged to empower themselves in accordance with some personal essence, each rational agent has an obligation to empower themselves in accordance with the impersonal essence of rational agency itself, i.e., to become more free by becoming better rational agents. In essence, reason demands of us that we unshackle it from the limitations of the form in which it is instantiated.

It’s important to note that, much as with Spinoza, this is not all there is to the ethical for Foucault. To complete my self-quotation:-

However, it should not for that matter be taken as an alternative to act-centred ethics. Foucault is not calling for the collapse of social norms (be they ‘ethical’ or otherwise) into the play of self-development. This is the import of Foucault’s distinction between ethics (norms of self-relation) and morality (norms of other-relation), and although he is certainly critical of the contemporary tendency to ignore the former in favour of the latter, this does not mean that he wishes to reverse this trend.

The problem is that Foucault never completed his account of how these two normative dimensions (self-relation and other-relation) are integrated. However, there are certainly suggestions that can be taken up from his work, the most clear of which seems to be that the ethics of self-relation are to be supplemented by a politics of other-relation. What Foucault calls morality, far from being eliminated, is to be transposed from the religious to the political sphere. The parallel with Spinoza is rather obvious here. The question that faces us is how precisely to carry out this transposition within the context of Foucault’s ethical innovations.

An interesting way of going about this is to see how we could translate Spinoza’s mereological solution to the problem into the Kantian framework we’ve been outlining. The important thing to recognise here is that Spinoza’s ontology contains no monads (or simples) but only composites. All modes are constituted by other modes, and this proceeds all the way down to infinity. Even if we’re sympathetic to this ontological position (which I am), the shift from ontology to deontology reintroduces a distinction between deontological monads and deontological composites. This is to say a distinction between individual rational agents, who are not composed of other rational agents (e.g., my heart, lungs, and liver cannot be functionally described as having the capacities necessary to count as rational agents), and properly collective rational agents, who are composed of other rational agents (e.g., the British government is composed of individuals who are themselves rational agents). To accept that there are such genuinely collective rational agents is to develop the Kantian account of subjectivity just outlined into a Hegelian account of spirit.

It’s worthwhile unpacking this idea in a bit more detail. It is crucial to remember the abstractness of the functional structure of rational agency discussed above. If it doesn’t matter what material this functional structure is instantiated in, it is entirely possible for it to be instantiated by a group of other rational agents, as long as they are related in the appropriate ways. For example, taking the schema of reasoning, perceiving, acting and coping presented above, we can understand a democratic state as being composed of a central political discourse (e.g., a parliament) that fills the role of the CRS, connected to various actional subsystems (e.g., military, police, etc.) via its executive (e.g., the government), receiving inputs from various scientific discourses and other data collection and processing mechanisms that play the role of perceptual subsystems (e.g., scientific advisory bodies, think tanks, etc.), all situated in the context of a plethora of institutions that carry out its ordinary functions, playing the role of coping subsystems (e.g., the civil service, the tax collection agency, local councils, etc.). It’s important to bare in mind that a state doesn’t have to be a very good rational agent in order to count as one. The functional nature of the structure of rational agency allows us to make normative assessments of how effectively any particular structure implements it, and so we can castigate democracy in the UK for being fundamentally unable to deploy the best scientific data it receives within its reasoning, unable to effectively work out contradictions in the various policy commitments it undertakes (practical commitments) and the rationales it provides for them (theoretical commitments), and being essentially akratic (i.e., weak-willed) in it’s ability to translate its commitments into action.

Nonetheless, it’s still important to insist that not every group of rational agents who form some kind of larger causal system thereby form a collective rational agent. A mob may have something like a causal unity wherein the actions of the various individual agents that compose it react upon one another to constitute emergent tendencies and capacities, but this doesn’t make it an agent in its own right. Foucault is famous for warning us of the dangerous temptation to interpret any kind of social structure as if it were a single agent, amenable to description in simplistic intentional terms. His account of Power is supposed to provide us with a set of tools for providing alternative functional descriptions of the sort of decentralised networks of social interactions that constitute the majority of the social field. These practical systems are not all collective agents, even if all collective agents are practical systems.

When we try to use this Hegelian view of deontological composition to reconstruct Spinoza’s mereological ethics, we encounter an interesting tension between the parts and the whole. Although it’s not possible for a group of rational agents to compose a collective agent without being configured in an appropriate way, its important to realise that this configuration need not have their consent. Precisely what does and does not constitute consent (and thus complicity) is a tricky issue, but we can nonetheless recognise that it is possible for individuals to play roles within systems of collective agency to which they have not committed themselves to in any reasonable fashion. Slavery is the most obvious example here, but there are plenty more banal examples of emergent systems of rational action to be found in the world of the modern workplace, wherein the parts are for the most part reluctantly involved. This demonstrates a tension between the freedom of the individual agent qua part and the freedom of the collective agent qua whole. If the collective rational agent is simply supposed to maximise its own positive freedom, then there is no reason for it not to do so at the expense of the freedom of its parts. Indeed, Taylorism advocates the limitation of the freedom of the parts as a means toward the maximisation of the freedom of the whole (see Zamyatin’s We as the classic thought experiment on this issue).

The problem here is that we have not yet uncovered anything like an essence of the state which would impose limitations upon its process of self-realisation. Without this, it is governed by precisely the same formal end as individual rational agents, namely, striving for individual autonomy. What is required is that this be substituted with some notion of collective autonomy, in which the freedom of the parts is not subsumed within the freedom of the whole. The best way to do this is to show that the latter is a means in relation to the former as an end, or that the power of the collective as a rational agent in its own right is an enabling condition of the power of the individual rational agents that compose it. There are two distinct dimensions to this: the role of the collective in providing capacities that its members would lack without it (e.g., the way a union makes possible collective bargaining, or the way that a state can make possible contractual relations that otherwise would be unenforceable), and the role of the collective in resolving conflicts between the freedoms of its members. We will call these the ampliative and integrative dimensions of collective agency, respectively. The subordination of collective freedom to individual freedom means that we are only justified in limiting the freedom of individuals in order to increase the freedom of the collective if this in turn increases individual freedom in another way. The collective essentially mediates the trading of different individual freedoms against one another.

What this does is to create a sort of scale invariant ideal regarding the relationship between the freedom of the parts and the freedom of the whole. Obviously, the freedom of the deontological monads is what underlies everything else. This isn’t necessarily to say that there isn’t any value to the freedom of the deontological composites apart from its ampliative and integrative function, but it is obviously subordinate to the value of monadic freedom. There can then be as many different mereological levels as you like (e.g., family, city, state, federation, etc.), and the relationship between the intermediate layers of deontological composites will be the same, forming a neat hierarchy: the freedom of the level above is subordinate to that of the level below, bottoming out in monadic freedom. I think that this ideal is what we would call Justice. It is the notion of collective autonomy in accordance with which we carry out the political process of constructing the state as a work of art, corresponding to the notion of individual autonomy in accordance with which we carry the ethical process of constructing ourselves as a work of art.

4. Conclusion: A Platonic Law

We now have the basis of an agent-centred politics that complements Foucault’s agent-centred ethics, though what is perhaps more interesting is the way it parallels the central idea of Plato’s Republic: Justice is the mental health of the state. The state considered as a collective rational agent is governed by the same functional norms as individual rational agents, even if these are supplemented by considerations regarding the balance of capacities between parts and wholes. We need not accept Plato’s picture of the mind, or his conviction that we can deduce a perfect state from this picture, in order to see the truth of this idea. Justice is not an archetype of the state, but an ideal that takes the form of a task: the task of experimentally constructing social relationships with the aim of achieving a state of increasingly harmonious freedom.

There are still a few issues that need addressing, one of which is appropriately Platonic in origin. We still have not properly dealt with the issue of consent, or the problem of self-determination we posed for Spinoza. This is demonstrated by the fact that Plato’s Republic would potentially be permissible on the account provided so far. What I mean by this is that it would be possible to construct a state wherein the freedom of political self-determination, or the right to involvement in the political process of some form, was traded off for increased freedoms of other kinds. Some might see no problem with such a benevolent dictatorship, but I think that it can (and should) be ruled out in principle. The crucial issue is the principle of autonomy mentioned earlier (also discussed here). This is the idea that we are only committed to things that we somehow commit ourselves to. The important word in the last sentence is ‘somehow’, because it isn’t necessarily the case that we have to understand precisely what we’re committing ourselves to in order to commit ourselves to it. The important thing is that if one is to be somehow bound by the norms that are constitutive for particular collective, then there needs to be a story explaining how one becomes so bound. This is to say a story about how one becomes a genuine member of that collective.

This is one more thing that I can’t provide a complete account of, as there are some rather tricky details to work out. The biggest difficulty is explaining how we become part of communities that we don’t choose to join, but are merely socialised into. This is particularly pressing insofar as membership in such a community is what binds us to the norms governing the process of undertaking commitments themselves. I’ve often talked of the transcendental norms of rationality, which are the universal ahistorical norms governing the game of giving and asking for reasons as such. I’ve also said that these present a sort of limit-case of autonomy, because we are committed to them insofar as we are committed to anything at all. However, there are what we might call institutional norms of rationality, which are particular historically instituted norms that instantiate the transcendental norms (I talked about these under the heading of ‘structural norms’ here). The point here is that there need to be specific norms governing our practices for individuating what counts as a rational agent, and what counts as an act which commits them to something (be it a theoretical or a practical commitment). These include the norms of the language we use (e.g., which sentences can be inferred from other sentences), as well as other basic norms of social interaction (e.g., how to query or challenge someone else’s commitments), and even norms governing the socialisation of new rational agents (e.g., how to treat children who are not yet fully fledged rational agents).

The relation between the transcendental norms and the institutional norms is analogous to that between a protocol and its implementation. No specific implementation of the protocol is strictly necessary, as it could always have been implemented in other ways, but it is necessary that there be some implementation of it. Similarly, no particular set of institutional norms is necessary, but it is necessary that there be some such norms. One is bound by these norms in virtue of one’s socialisation into the particular rational community these norms institute, though one can become bound by others insofar as one enters into rational discourse with another such community. These institutional norms are thus part of the limit-case of autonomy. We are committed to them insofar as we use them to commit ourselves to anything. We are thus a member of a rational community insofar as we engage within others rationally within some instituted context. This is the most fundamental (and minimal) form of rational collectivity on top of which everything else is built. One can then be socialised into, implicitly incorporated into, or explicitly commit oneself to, other more substantial communities, such as societies that are built on laws. However, these exist outside of the limit-case of autonomy, and thus require something beyond the mere act of committing oneself to something to be bound by them. This means that one must have a say in how one is bound, no matter how slight. I think this is what ultimately undermines the dictatorial state, no matter how benevolent it may be.

The crucial thing here is that the ideal of Justice confronts all of us with the same task, namely, constructing the rational communities we are part of along certain lines. This means we are confronted with the task of instituting norms that structure the community in such a way as to amplify and integrate the freedoms of all those who are involved. This is to say that we are confronted with the task of instituting Law. It is important to remember here that not all norms are laws. Dinner party etiquette, the rules of football, the fashions of high society clothing, all these are in no sense legal norms. It is only those norms which attempt to instantiate Justice that are properly called laws. It is equally important to remember that legal institutions can take many different forms. Case law, the separation of legislature and judiciary, and the trial system are all contingent particulars of the way Law is instituted. What is universal is the fact that we all face the Law as equals, not in the manner of Kant’s moral law, wherein we must be able to universalise our own actions, but in virtue of the fact that we all are confronted with the task of constructing a free community. We are equal in the eyes of the Law, not principally as those who are subject to it, but as those who create it. It is thus entirely possible to institute differing legal statuses that grant different rights and responsibilities to those who occupy them. It may be permissible for a policeman or soldier to perform actions that don’t pass Kant’s test (at least crudely understood), as long as the role he occupies is part of the design of a Just society. I take this to be the essential insight of Rawls’ original position.

The most important thing about Law is that it is always in tension with Justice. The laws we institute are our attempts to instantiate Justice, but they can be imperfect or be rendered invalid by changes in our social conditions. This means that Law is always subject to challenge in the name of Justice. The authority of laws is not a matter of fiat. They are not simply commandments that are past down from on high, but are rather subject to challenge, reinterpretation and revision on the basis of reasons. If legal institutions do not take this into account then they are fundamentally unjust and therefore not really legal institutions. The Law must incorporate the means for its own revision, and these means must in some sense be open to all. It is when these means are truncated, distorted, or otherwise removed that it becomes possible to break the Law in the name of Justice, in order to reinstitute Law on a different basis. This is what undermines the benevolent dictatorship: freedom without the possibility of involvement in the process of instantiating Justice can never itself be Just. This means that there will always be circumstances in which we are justified in breaking with the norms of any rational community we have been bound to, in the name of Justice.

This returns us to Spinoza, and the topic we started out from: the elimination of norms in favour of causal knowledge. I believe that the account I have just outlined incorporates the real promise of Spinozism, namely, the eminently rationalist demand that all of our ethical prescriptions be based on good reasons. At no point should questions of the form ‘why should we (not) do x?’ receive answers of the form ‘because Y says so’. Moreover, this has been done in a way that retains the appeal to causal knowledge through the use of specifically instrumental reasoning. However, this instrumental reasoning is reasoning regarding what laws we should institute, so as best to achieve the end of maximising our collective freedom. It thus can’t be represented as anything like an elimination of norms. It is instead the call for a maximal involvement of our causal understanding of ourselves (both as individuals and as collectives) in the process of legislating the principles we are to live under, with the aim of building of and for ourselves a Just society. If anything, this supplemented Spinozism demands of us a renewed concern with norms, and the practices through which we institute them.