Reasons Without Persons is the first book-length development of the increasingly influential ‘time-slice’ approach to rationality, which holds that rationality is purely synchronic – that the fundamental rational norms are concerned only with the way we are at particular moments, and not the way we behave over time. Those familiar with Hedden's existing papers on this topic will not find much new or surprising here, but it adds something to see how all the parts of his picture hang together, when elegantly compiled. Even to someone unsympathetic to its overall thesis, as I am, the book is insightful, admirably clear, well-structured, well-written and generally well-argued, and serves as an excellent point of embarkation for anyone interested in exploring the relationship between rationality and time. Though the book is especially concerned with the sorts of principles developed by formal epistemologists and decision theorists, it is not particularly technical, and its implications are quite general.
Hedden's conception of the time-slice project has two main commitments – first, that an agent's rationality does not depend on their past mental states, and second, that an agent's rationality does not depend in any special way on their beliefs about their past mental states. He begins, in chapter 2, by providing some general motivations for these claims, and then proceeds for the bulk of the book to turn attention to two types of norms from formal epistemology and decision theory that the time-slice picture must reject – Conditionalization and Reflection.
Hedden addresses these methodically – first arguing directly against them as constraints on belief and then against their analogues for preferences (chapters 3–4). Next, he addresses the main positive arguments for those principles, those appealing to diachronic tragedy, where agents perform sequences of actions which are clearly undesirable, but where each action individually looks sensible by the lights of the present time-slice. He defends a theory of options – the candidates for what one ought to do – as the decisions an agent can make, and shows how appreciating this theory undermines the apparent threat posed by diachronic tragedy (chapters 5–7). Finally (chapters 8–10), he tries to show how a system of purely synchronic, time-slicer-friendly replacement principles can fill the gaps left behind by rejecting diachronic norms (or at least show that where gaps remain, they need not be filled).
There is much to say about each part of this dialectic, but I want to focus instead on an issue lurking in the background throughout the book – the extent to which a theory of rationality ought to idealize away our cognitive limitations. I think there is a tension in the way Hedden treats this issue, one that I think points to an attractive but underexplored direction for resistance to the time-slice picture.
When Hedden is on the offensive, he is not shy about raising worries that his opponents are idealizing too much. He objects to conditionalization for implying that forgetting is irrational (p. 43), and dismisses the idea that an account of ideal rationality can abstract away from our actually imperfect memory. Later (p. 127) he raises a similar point against views that require an ideally rational agent to have the ability to self-bind – to exercise causal power now over their decisions in the future.
I think these are apt criticisms. But Hedden's own view runs into worries about over-idealization as well, and his defences are not very compelling. There are two related concerns. First, the time-slice view does not allow a rational agent any lag between gaining evidence and fully updating their attitudes, something which in ordinary humans always takes at least some (and often considerable) time. Second, the time-slice view does not supply any rational requirements on reasoning from the evidence, something which only occurs over time and yet seems paradigmatically rationally assessable. Hedden's rational agents update all their attitudes instantaneously without the need for any kind of reasoning. On the face of it, this seems like a very extreme idealization. We do not typically take humans to be irrational merely in virtue of being the sorts of beings that need to engage in reasoning or that take non-zero time to form beliefs.
Hedden gives two kinds of response. First, he suggests that it's okay to abstract away from the fact that updating takes time because the resulting theory is simpler than any theory which takes these cognitive limitations into account (p. 148). He goes so far as to propose (though not defend) that being simple is part of what makes it the correct theory of rationality, but I think this appeal to simplicity is troubling even on a merely methodological reading.
One problem is that in order to explain things like the irrationality of highly unstable belief, Hedden must supplement the basic synchronic picture with something like Uniqueness, the claim that there is only one doxastic state it is rational to have on any given evidence. Any account of how to match a set of evidence with the unique credence function it supports will plausibly itself be very complex (Hedden himself does not even attempt to provide one). But stability is something diachronic views of the sort he criticizes can accommodate without any such auxiliary theory. So if it is fair for Hedden to appeal to simplicity as a trump card here, it looks fair for his opponents to make the same appeal in response to his own challenges.
More importantly, however, while simplicity is a virtue in a theory, it must not come at significant cost to the things the theory was supposed to explain. And rationally requiring agents to do things they cannot do – like update their attitudes instantaneously – threatens the adequacy of Hedden's view both as an account of our pre-theoretic notion of rationality and for the things he himself identifies as its main theoretical roles. A theory of rationality, Hedden thinks, should (a) provide grounds for criticism of agents, (b) play a major role in predicting and explaining the behaviour of agents, and (c) provide guidance to agents. But it is inappropriate to criticize people for failing to update instantaneously when this is something they lack the capacity to do. A theory of rationality will be of very limited use in predicting or explaining our behaviour if rational agents always update instantaneously without reasoning. And agents like us who need to update our attitudes through reasoning plausibly need guidance about how to reason.
Later in the book, Hedden takes a different approach. The requirements of rationality, he suggests, are necessary – they must apply to any rational being, including those without our contingent psychological limitations (p. 186). Consequently, a requirement that applies to a superior being such as an angel applies to us as well. And since it would be irrational for an angel with none of our limitations to take time to reason or update beliefs, it would be irrational for us to do it as well.
But this argument makes a serious mistake. Compare Hedden's argument to the following (obviously bad) parody: the requirements of morality are necessary. So, a requirement that applies to Superman also applies to us. Superman is morally obligated to leap off a building if he sees a crime being committed on the street below. So, we are morally obligated to leap off a building if we see a crime being committed on the street below.
The error in this argument is that what is necessary about morality are the most general and fundamental requirements, not the particular prescriptions, which are contingent and depend on the (in this case, physical) abilities of particular moral agents. The same goes for rationality. In order for a theory of rationality to respect our cognitive limitations, all we need is for particular prescriptions to differ depending on the agent's capacities. And this is consistent with the general requirements themselves being necessary. For example, the fundamental and necessary requirement may be something like ‘update your beliefs as fast as you can’. This will lead to angels being required to update instantaneously and agents like us being permitted to take some time. Nothing about this violates any plausible necessity constraint.
It is notable that the diachronic norms most prevalent in the literature and which Hedden spends the bulk of his time addressing, like Conditionalization, are themselves highly idealized, and so are ill-poised to take advantage of the problems time-slice rationality has with respecting our cognitive limitations. I think it is not coincidental that it is when engaging with these norms that Reasons without Persons is at its strongest. This suggests to me that philosophers may have been looking in the wrong places for diachronic norms by relying on the idealizing frameworks of formal epistemology and decision theory.
These issues aside, Hedden's challenges are generally forceful, and I believe anyone interested in defending a diachronic picture of rationality should be prepared to address them. Time-slice rationality deserves to be taken seriously, and this is the clearest, most complete and most persuasive exposition of the approach to date.
