Rational Powers in Action
Instrumental Rationality and Extended Agency
SERGIO TENENBAUM
Great Clarendon Street, Oxford, OX2 6DP, United Kingdom
Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries
© Sergio Tenenbaum 2020
The moral rights of the author have been asserted
First Edition published in 2020
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above
You must not circulate this work in any other form and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2020937535
ISBN 978–0–19–885148–6
Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY
Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.
Preface
While on the subway to work I space out and, before I know it, I’ve reached my destination. But there were many things I could have done between the time I boarded the subway and my final stop. At each moment, I could have chosen to grade a paper from my bag, or to read the fiction book that I downloaded to my tablet, or play some electronic games on my phone. There were also slight improvements that I could have made to my seating arrangements—I could have gone closer to the door, or away from the noise bleeding from my neighbour’s headphones—improvements that I could have weighed against the inconvenience and effort of moving from one seat to another.
On many views of instrumental rationality, especially those that take orthodox decision theory as their starting point, my failure to consider these options, or at least to act on them if they would turn out to be preferable, shows that I have fallen short of ideal rationality. Of course, everyone will rush to add that given our limitations, we should not engage in trying to maximize utility at every moment; we’re better off using heuristics, or tried and true strategies that allow us to approximate ideal decision making as much as possible given our limited rational capacities. On this view, an ideally (instrumentally) rational agent chooses the best option at each moment at which she acts. I argue in this book that this is a fundamentally flawed picture of rational agency given the structure of human action.
We are often (indeed, arguably always) engaged in actions that stretch through extended periods of time in the pursuit of less than fully determinate ends. This basic fact about our rational existence determines a structure of rational agency that is best captured not in terms of the evaluation of moment-by-moment choice, but rather by the (attempted) actualization of various ends through time. The point is not that, given the vicissitudes of the human condition (that it takes time to deliberate, that our calculating powers are modest, and so forth), it is too demanding to evaluate our choices in terms of moment-by-moment maximization of utility (let alone to enjoin agents to be explicitly guided by an ideal of moment-by-moment maximization). Rather, given the nature of what we pursue and how we pursue it, a theory of moment-by-moment maximization, or any understanding of
instrumental rationality on the basis of momentary mental states, cannot capture the fundamental structure of our instrumentally rational capacities. Instrumental rationality is rationality in action, and, in particular, in action that extends through time.
The book aims to provide a systematic account of the nature of instrumentally rational agency in the pursuit of long-term, not fully determinate ends; that is, ends that cannot be realized through a single momentary action and whose content leaves partly open, at least to the agent herself, what counts as realizing the end. The restriction of the scope of the theory to these kinds of ends does not represent a significant narrowing down of the realm of rational agency; at least as far as human agents are concerned, most, if not all, of the cases in which someone exercises her capacity for rational agency are instances of rational agency in the temporally extended pursuit of indeterminate ends.
Chapter 1 presents what I take to be the general structure of a theory of instrumental rationality, and lays down some of the main ideas and motivations for the extended theory of instrumental rationality (ETR) developed in the book. I characterize the content of a theory of instrumental rationality in terms of its given attitudes (the “inputs” for a theory of instrumental rationality), its principles of coherence, its principles of derivation, and what counts as the conclusion of practical reasoning or the exercise of our powers of instrumental rationality (what the theory of instrumental rationality takes to be the “outputs” of practical reasoning). I then provide a way of distinguishing the domain of instrumental practical rationality from the domain of substantive practical rationality. Chapter 2 starts presenting the main tenets of, and the initial motivation for, ETR. According to ETR, both the given attitudes (“inputs”) and the conclusion of practical reasoning (“outputs”) are actions (more particularly, the intentional pursuit of ends). The principle of instrumental reasoning, which tells us (roughly) to adopt means to our ends, is the only principle of derivation, and a principle (roughly) requiring agents not to pursue incompatible ends is the only principle of coherence. The chapter also presents one of the main theses of ETR; namely that the rationality of an agent through an interval t1−tn does not supervene on the rationality of the agent at each moment t1−tn (what I call the “non-supervenience thesis”). In other words, someone may be irrational over a period of time without there being any moment during that time at which they were irrational. The main contrast to ETR in Chapter 2 is with a version of orthodox decision theory; contrasting the view this way help us understand some of the most important potential advantages of ETR.
Chapter 3 continues developing ETR, and starts the proper positive argument for the theory. ETR takes the intentional pursuit of ends to be the only relevant attitude for the theory of instrumental rationality, and takes the principle of instrumental reasoning to be the only principle of derivation. However, isn’t practical rationality essentially comparative? Doesn’t a rational agent choose the best alternative among a set of options? In particular, it seems that if the agent has more than one end, we’ll need to introduce comparative or graded attitudes, such as the preference orderings in orthodox decision theory, in order to explain the rationality of her choices in contexts in which the agent needs to choose among competing ends. The chapter argues that the lack of comparative or graded attitudes does not prevent ETR from providing an adequate account of rational agency in contexts in which an agent has multiple ends that can be distinctly pursued through different actions. In fact, ETR does better than theories that rely on comparative and graded attitudes in accounting for the fact that, given our pursuit of indeterminate ends through time, in many, if not most, situations, there is no “best” option. I argue that in these situations our only option is to “satisfice”. At the same time, the chapter explains how some comparative attitudes, such as preferences, can be incorporated into ETR in specific contexts.
Chapter 4 looks at Quinn’s puzzle of the rational self-torturer in order to establish the truth of the non-supervenience thesis. The puzzle presents, in a rather simple and stark way, a structure that pervades our pursuit of ends through time. I argue that that a proper solution for the puzzle (and thus a proper account of instrumental rationality that applies to extended action) must accept the non-supervenience thesis. But given that long-term actions are pursued through momentary actions, we need to understand how the agent’s extended perspective (the perspective of the pursuit of long-term ends) and the agent’s punctate perspective (the perspective of the pursuit of momentary actions) interact in realizing the agent’s indeterminate ends. This chapter presents ETR’s account of this interaction. Since extant theories of instrumental rationality cannot do justice to the non-supervenience thesis, and a fortiori, to the structure of agency illustrated in the puzzle of the rational self-torturer, this chapter represents an important argument in favour of ETR.
Chapters 5 and 6 consider whether introducing future-direct intentions in our account of instrumental rationality can do a better job than (or at least complement) ETR in accounting for extended agency. Work on the relevance of future-directed intentions to rationality would seem particularly
promising in this context, as future-direction intentions are often introduced exactly to do justice to the temporally extended nature of our agency. I argue in these chapters that attempts to rely on future-directed intentions can do no better than ETR in explaining rational agency through time and often end up generating spurious requirements. I argue that these theories try to understand the structure of rational agency through time as a diachronic structure rather than a properly extended structure, and therefore these attempts are bound to fail.
Most theories of rationality take it for granted that rationality is always a matter of conformity to certain principles, or responding to reasons. In Chapters 7 and 8 I argue that (non-trivial) principles of rationality cannot fully capture the nature of our instrumentally rational powers. There are also what I call “instrumental virtues”. An agent suffers a defect of rationality insofar as she fails to have one of these virtues, or insofar as she manifests one of its corresponding vices. However, an agent can manifest instrumental vices without violating any principle of rationality. Chapter 7 proposes that courage is one of the instrumental virtues, and shows how one of its corresponding vices, cowardice, must be understood as a defect of our rational powers. Chapter 8 argues that ETR gives us the tools to account for an overlooked instrumental virtue, which I call “practical judgment”. An agent exhibits the virtue of practical judgment insofar as she can pursue long-term ends without relying too much on restrictive implementation policies. The account also allows us to understand ordinary phenomena such as procrastination as manifestations of practical judgment’s corresponding vices.
Chapters 1 through 8 proceeded mostly assuming that the agent has all the relevant background knowledge that she needs to pursue her ends. But this raises the question of whether the theory can be extended to contexts of risk and uncertainty. This is particularly important given that orthodox decision theory provides a powerful account of the rationality of choice under risk. Chapter 9 argues that ETR can provide a satisfactory account of rationality in these contexts and that it can actually co-opt the resources of decision theory exactly in the cases in which the theory seems most plausible; namely, the pursuit of what I call “general means” (such as the pursuit of health or wealth). Moreover ETR plausibly renders coherent certain ubiquitous choice dispositions that seem incompatible with orthodox decision theory.
Many people have helped me in the course of writing the book and developing these ideas; many more than I can presently name. But I would certainly like to thank for their invaluable feedback, Chrisoula Andreou,
David Barnett, Christian Barry, Michael Bratman, Mark Budolfson, Ruth Chang, Philip Clark, Brendan de Kenessy, Matthias Haase, David Horst, David Hunter, Thomas Hurka, Douglas Lavin, Erasmus Mayr, Julia Nefsky, Sarah Paul, Juan Piñeros Glassock, Douglas Portmore, Diana Raffman, Arthur Ripstein, Karl Schafer, Michael Thompson, Benjamin Wald, and Jonathan Weisberg. Alice Pinheiro Walla organized a workshop at Bayreuth for the book manuscript that not only pushed me to complete a draft of it, but also made me rethink many of the ideas of the book. I am immensely grateful to her and the other commentators at the workshop: Franz Altner, Luca Ferrero, Erasmus Mayr, and Franziska Poprawe. Luca Ferrero and Jonathan Way refereed the manuscript for OUP, and their thoughtful and insightful comments made me rethink many of the book’s arguments, and greatly helped making them clearer, and hopefully, more compelling. Peter Momtchiloff at OUP was amazingly supportive and helpful through the entire process. I feel truly fortunate to have worked with him.
Many ideas of the book were developed and discussed in the annual workshop meetings of the Practical Thought and Good Action Network (later Action Network). These events were models of philosophical discussion and interaction, and I greatly benefitted from all our meetings. I would like to thank all participants and especially Matthias Haase, who kept the network going throughout the years.
I started worked on the book in the 2011–12 academic year. I received a fellowship for the Institute for Advanced Studies at the Hebrew University of Jerusalem for the Fall term of 2011, and I was a fellow at Magdalen College at Oxford University during the Trinity term in 2012. I made the final substantive revisions during the Summer and Fall of 2019, when I was a Research Fellow at the Australian National University (Summer) and received a fellowship from the Jackman Humanities Institute at the University of Toronto (Fall). The research on this book was also made possible by grants from the Social Science and Humanities Research Council of Canada (SSHRC). I am very grateful to all these institutions for their generous support. Earlier versions of some of the material in this book has appeared elsewhere. An earlier version of parts of Chapter 3 appeared as “Acting and Satisficing” in Legal Normativity and the Philosophy of Practical Reason, edited by Georgios Pavlakos and Veronica Rodriguez-Blanco (Cambridge University Press, 2015), 31–51. Earlier versions of Chapter 4 appeared as “Vague Projects and the Puzzle of the Self-Torturer” (with Diana Raffman), Ethics 123 (October 2012): 86–112. An earlier version of Chapter 5 appeared as “Reconsidering Intention”, Noûs 52 (June 2018): 443–72. An earlier
version of Chapter 6 (except for Section 6.5) appeared as “Minimalism about Action: A Modest Defence”, Inquiry (special issue on Choice Over Time) 57 (April 2014): 384–411; an earlier version of Section 6.5 appeared as “SelfGovernance Over Time”, Inquiry (September 2019): 1–12. Finally, an earlier version of parts of Chapter 8 appeared as “The Vice of Procrastination” in Chrisoula Andreou and Mark White (eds), The Thief of Time (Oxford University Press, 2010), 130–50. I am very grateful to the feedback provided by all of those I have mentioned in these publications. I would also like to thank here the various audiences that provided invaluable help with various versions of “Instrumental Virtues and Instrumental Rationality”, which makes up the bulk of Chapter 7 and has not previously appeared in print: the material there was greatly improved by the feedback I received from audiences at Australian National University, Florida State University, the Halbert Network Meeting at the Hebrew University of Jerusalem, Universidade de Caxias do Sul, University of Michigan, the Workshop on Normativity and Reasoning at NYU, Abu Dhabi, and my commentator on this occasion, Eric Wiland.
My sons, Alexander Tenenbaum and Leonardo Tenenbaum, have put up with their father asking them bizarre questions about practical rationality on way too many occasions. People unavoidably judge their children to be exceptionally wonderful human beings, and I feel so fortunate that in my case this is actually true. My greatest debt is without a doubt to Jennifer Nagel. If any thought in this book is worth anything, it is because it has been improved beyond recognition through Jennifer’s incomparable wisdom (and patience!). Moreover, if you write about ideal agency, it cannot but help to live with someone who so closely approaches this standard.
1 Extended Action and Instrumental Rationality
The Structure of a Theory of Instrumental Rationality
1.0 Introduction
The aim of this book is to provide a systematic account, or at least the beginnings of a systematic account, of the nature of instrumentally rational agency in the pursuit of long-term, not fully determinate ends; that is, ends that cannot be realized through a single momentary action and whose representation leaves partly open, at least to the agent herself, what counts as realizing the end.1 The restriction of the scope of the theory to these kinds of ends does not represent a significant narrowing down of the realm of rational agency that it covers. Nearly all, and arguably all, of the cases in which human agents exercise their capacity for rational agency are instances of rational agency in the pursuit of indeterminate ends through an extended period of time. Thus the more ambitious aim is to provide a theory that encompasses the entire domain of instrumental practical rationality, or, more precisely (and moderating the ambition somewhat), a theory of the basic foundations of instrumental practical rationality.
Suppose, for instance, today is Monday and I am engaged in making the house look nice for a visit from my aunt, who will be arriving some time Wednesday afternoon. My having this end leaves various things indeterminate. Should I hang paintings on the wall, or put flowers at her bedside, or both? At some point, I might form a more determinate conception of what counts as making the house nice for my aunt, but it seems that I can
1 Of course, this is not supposed to be a very precise characterization of these notions. I’ll start by relying on an intuitive understanding of these notions, and try to provide more precise characterizations as the book progresses.
Rational Powers in Action: Instrumental Rationality and Extended Agency. Sergio Tenenbaum, Oxford University Press (2020). © Sergio Tenenbaum. DOI: 10.1093/oso/9780198851486.001.0001
have the end of making the house look nice for my aunt without having settled this question. Similarly, what counts as achieving this end can be vague in many ways. How nice is nice enough for my aunt? If there is a small corner of my bedroom that has some dust in it, does it count as being nice enough? How much dust in the house will be enough to determine that the house is not nice? At precisely what time should everything be ready?
Moreover I can’t just make the house look nice in one magical momentary twitch of my nose. I have to engage in various actions at some point between now and Wednesday afternoon, but there are (almost) no specific points in time at which I will be presented with a precise set of options such that I must choose a certain alternative if I am to make the house nice for my aunt. Everything (or at least nearly everything) I do could have been done slightly later, with slightly less effort, and so forth. Given these facts about the indeterminacy of my project and the way it stretches over time, what exactly am I rationally required to do at various moments or intervals in light of this project? How do the requirements generated by this end interact with requirements generated by other ends that I am pursuing at the same time (I am also writing a book, I need to teach Tuesday evening and be prepared for it, etc.)? What would be the virtues of character of ideally rational agents (and what would be the corresponding vices in less ideal agents) who can efficiently pursue ends of this kind?2 This book will try to answer these questions and related ones.
On the view I defend here, instrumental rationality is, roughly, a relation between intentional actions. More specifically, it is the relation of the pursuit of some actions as a means and the intentional pursuit of an indeterminate end extended through time, in which the latter explains the former in a particular way. So my writing this sentence is a means to the intentional pursuit of writing this book, while writing a book (and also writing this sentence) is something that I pursue over an extended period of time. Moreover, the end of writing a book is indeterminate; in pursuing the end of writing a book, I do not (necessarily) specify the precise quality, length, or completion time, even though not all combinations of quality, length, and completion
2 It is worth noting that indeterminate ends are not like “gaps” in a preference ordering. Gaps seem to imply that there’s something missing there; an agent with a complete preference ordering has, at it were, made up her conative mind about a larger number of things than the agent with gaps in her preference ordering. But there’s nothing missing in an indeterminate end; though we can talk about acceptable and non-acceptable determinations of an indeterminate end (and, as we will soon see, distinguish different ways of specifying the end), a determinate end is a different end, not a more “complete” one.
times are acceptable realizations of my end.3 Note also that “pursuing an end” is not essentially graded or comparative. That is, pursuing an end does not come essentially in degrees. You do not pursue an end to a greater or lesser extent; you are either pursuing it or not. Or at least, I will not rely on any assumptions that ends are pursued to different extents; if there is sense to be made of the idea of “graded pursuits”, it will be irrelevant for our theory. In contrast, on some views of beliefs and desires, beliefs and desires come in degrees, and the theory of rationality depends not only on the content of these attitudes, but also on their degrees. Similarly, the central category for my theory is “pursuing the end E”, rather than “pursuing an end E* to a greater extent than E#” or “pursuing and end E* over E#”. In contrast, a number of extant theories of rationality give attitudes that are essentially comparative, such as preferences, a starring role.
More specifically, the opposing views take instrumental rationality to depend on (i) comparative or graded attitudes, and to concern (ii) relation between mental states that are (iii) momentary, and (iv) whose objects are, in the relevant way, determinate. Not all opposing views need to accept all four parts of this claim. I’ll start in the first two chapters contrasting the theory I’ll develop here, the extended theory of rationality (ETR) with orthodox decision theory. Orthodox decision theory is obviously an important and influential view, but, more importantly for our purposes, it contrasts in all these aspects with ETR, and so it is an ideal contrast with the theory.4
In Section 1.1 I give a very rough outline of the contrast between orthodox decision theory and ETR, as well as some initial motivation for ETR, focusing on the contrasts marked as (iii) and (iv).5 After this initial contrast, I move on to more structural questions. Sections 1.2, 1.3, and 1.4 introduce some important terminology and develop a general understanding of what a theory of instrumental rationality is. Section 1.2 aims to develop a neutral framework that can be used to compare and contrast different theories of instrumental rationality. Section 1.3 then puts forward some basic criteria for evaluating theories of instrumental
3 More on the notion of “acceptable realization” below.
4 Not all versions of decision theory fall into the category of a theory of instrumental rationality as I understand it here. So not all of them compete with ETR. More on this point below.
5 The contrast with (i) is discussed in more detail in Chapter 3 and (ii) in Chapters 4 and 5. Of course, since the book aims to give an account of instrumental rationality that generalizes to extended actions with indeterminate ends, the contrast with positions that accept (iii) or (iv) is discussed throughout the book.
rationality. Finally, Section 1.4 proposes an explanation of the difference between instrumental rationality and “substantive” rationality.
My approach to the theory of instrumental rationality is in some ways different from a number of contemporary approaches to the theory of instrumental rationality, so it is worth saying a few words about it at the outset.6 There has been a flurry of recent work on “coherence” and “structural” requirements of rationality and the normativity of the principles of rationality.7 Often, these philosophers will start with principles of coherence and ask questions about whether we have reason to conform to the principles, whether there’s value in a disposition to conform with them, and so forth. My starting point is the idea that we have certain rational powers and capacities to act, and the theory of instrumental rationality is the theory of a subset of these powers. The principles of rationality are thus the principles that, in some sense, explain the agent’s exercise of such powers. In the good case, a rational action is one that manifests this power. Cases of irrationality will be cases of failures to exercise the power, or improper exercises of the power. Of course, at this abstract level of description it is hard to see how this distinctive starting point makes a difference, and I hope the answer to this question becomes clearer as the book progresses. However, two points are worth noting. First, although questions about the normativity and the value of rationality are important, they are not the only ones. I think much is missed about the nature of instrumental rationality if we focus mostly on these questions, and taking the question about the nature of a rational power as our starting point allows us to circumvent these issues. This is, of course, a methodological point. There is no commitment here that the same views would not be accessible if we were to start somewhere else. Conversely, insofar as this book yields insights about the nature of instrumental rational agency, we have good reason to accept that taking rational powers or capacities as our starting point is a fruitful line of inquiry. Second, understanding rational agency in this way helps identify the kind of theory of rationality that I am proposing here. There are different types of theories of rationality. One can have a purely “evaluative” theory whose principles simply evaluate actions or mental states of the agent as rational or irrational, while making no claims about whether an agent is, or ought to be, guided by such principles. That is, the theory only speaks about
6 More on this approach in Section 1.4.
7 For a few recent examples of book-length works in this tradition, see Broome (2013), Kiesewetter (2017), Wedgwood (2017), Lord (2018), and Brunero (forthcoming).
principles in accordance with which a rational agent should form a mental state or act, but makes no claim about principles from which a rational agent ought to form a mental state or act. Though more common in economics, theories of rationality can also be purely descriptive: they can aim simply to explain human behaviour on the assumption that people by and large act rationally. Finally, a theory of rationality can be “action guiding”. Such a theory tries to describe the principles from which the agent acts insofar as the agent is rational.8 Since my interest is a theory that explains what it is to manifest these particular rational powers, it seems clear that the project falls roughly within the third type of theory of rationality. However, it is importantly related to the second: in the “good” case in which the agent’s exercise of her rational powers is flawless, the theory of rationality provides also the explanation of the agent’s behaviour. When a rational agent, for instance, believes Fa because she believes (x)Fx, the fact that Fa follows from (x)Fx fully explains the agent’s belief, and Universal Instantiation is the specific principle that explains why the agent believes Fa (together with the known fact that (x)Fx). And, similarly, cases of irrationality will be cases in which the agent failed to manifest this power when it could or when the manifestation was in some way defective (that is, when the resulting belief or other attitude cannot be fully explained in terms of the manifestation of a causal power).9 Here too the choice is methodological. I’ll not try to defend this conception of rationality in any detail;10 it’ll prove its value if it produces new insights in the theory of instrumental rationality.
1.1 Indeterminate and Long-Term Ends
During my morning subway ride, I was just daydreaming. But there were so many things I could have done between the time I got on the subway and the final stop. At each moment, I could have chosen to read a paper from my bag (which one? I brought my colleague’s book draft on Thales’ moral theory, the latest issue of The Philosophical Journal, and a student paper on Eurovision aesthetics). I could have read the fiction book that I downloaded to my tablet—although the tablet also had some electronic games that I
8 See Bermudez (2009) for a similar classification in the case of decision theory.
9 See Marcus (2012) for a very interesting attempt to provide an explanatory theory of action and belief in terms of rational abilities and their manifestations.
10 However, a possible advantage of this approach, given what I said above, is that it has the potential to provide the tools for unifying evaluative and explanatory theories of rationality.
could have been playing without feeling guilty about engaging in this vice during work hours. Between the third and fourth stop, there was a chance to glance over the headlines of the newspaper that the person in front of me was reading, and between the fourth and the fifth stop I could have struck up a conversation with the intelligent-looking fellow seated to my left. If I decided to start a conversation, there were many ways I could have proceeded: I could have started with the weather, or taken his baseball hat as a lead to talk about our hopes for the Blue Jays, or just dared to express my discontent with our current mayor. Moreover, I chose not to make stops, but this was not necessary; I have a subway pass and it would have cost me nothing to just get out at the first stop to buy bread at that lovely bakery, or to acquire some cheap comic books at the used bookstore at the second stop, or to get an espresso and reply to a few emails at a Groundless Grounds franchise at each station. It is not clear that any of these stops would compensate for the twelve-minute reduction in my planned eight-hour-and-sixminute workday, but I never even considered them.
Does my failure to consider these options, or at least to act on them if they would turn out to be preferable, really show that I have fallen short of ideal rationality.11 Of course, everyone will rush to add that given our limitations, we should not engage in trying to maximize utility at every moment; we’re better off using heuristics, or tried and true strategies that allow us to approximate ideal decision making as much as possible given our pathetically feeble rational capacities. Yet, we might ask in what sense and why this is an ideal of rationality. Let us say that the Vulcans realize this ideal; a Vulcan, say, always has a preference ordering that conforms with the axioms of decision theory, her choices always conform to these preferences, and her preferences have all the bells and whistles we’d want from a preference set (they are considered, well-informed, etc.). Why are the Vulcans more ideal than I am, at least as far as my subway ride is concerned (doubtless we’d find much to deride in my employment of my rational capacities in other parts of my life)?
One answer would be that they actualize more value, or respond better to reasons, than I do. This answer might be correct (though I don’t think so, or at least not necessarily so), but examining this type of answer is outside the
11 This claim (and some of what I say below) needs very many caveats. There are many interpretations of decision theory and its relation to normative theory. I discuss these issues in greater detail later in the chapter, so for now I offer the bold, though vague and possibly imprecise, unmodified claim.
scope of this book. It pertains to the substantive rationality of the Vulcans, rather than their instrumental rationality. I’ll discuss in greater detail below what counts as instrumental rationality, but for now, we can say that this alleged greater realization of value is not a difference in how the Vulcans succeed in realizing what they care about or what they’re trying to achieve; instead, it is a question about whether the Vulcans are pursuing the right thing. But even within the realm of instrumental rationality, it might seem obvious that the Vulcans are doing better: after all, they follow their preferences, and thus they expect to do better even when we take their own attitudes as the standard. In my subway trip I often failed to pursue my most preferred option, and thus to choose what I most wanted. So have I not failed by my own lights?12
However, this thought makes two assumptions, neither of which I think is ultimately correct. First, it assumes that I have the relevant preferences. It is rather implausible that I have determinate preferences for each possible choice situation that I can concoct. But, more importantly, I will argue that it is not true that an instrumentally rational agent always chooses the most preferred option. The main reason for this conclusion is that we often (arguably always) are engaged in actions that extend through arbitrarily long periods of time in the pursuit of less than fully determinate ends. This basic fact about our rational existence determines a structure of rational agency that is not best captured in terms of moment-by-moment choice, but rather by the (attempted) actualization of various ends through time. This structure will often permit, and at some level require, that sometimes I choose a less preferred option over a more preferred one. As we will see in Chapter 4, the point is not that given the limitations of the human condition (that it takes time to deliberate, that our calculating powers are modest, and so forth) it is too demanding to evaluate our choices in terms of moment-bymoment maximization of utility (let alone to enjoin agents to be explicitly guided by an ideal of moment-by-moment maximization). But rather given the nature of what we pursue and how we pursue it, a theory of moment-bymoment maximization cannot capture the fundamental structure of our instrumentally rational capacities. Roughly, my subway ride is rational, or as I prefer to say, an expression of my rational powers, because through this extended action I have non-accidentally actualized one of my ends (my arriving at work in time) without undermining any other end of mine.
12 To paraphrase Donald Davidson, I’d be losing by my own standards. See Davidson (1980b), p. 268.
Questions of moment-by-moment maximization do not even come up in accounting for my rational agency in this case. As will become clear later in the book, this is not to say that preferences and decision theory play no role in our understanding of instrumental rationality. Given the structure of some of our ends, and of some general means we use to pursue them, preference orderings will have an important role to play, even if a much more specialized and subordinate role than often assumed in the theory of instrumental rationality.
For orthodox decision theory, ideal rational agents have no indeterminate ends (preferences),13 and the rationality of their long-term actions supervene on the rationality of their momentary decisions.14 The axioms of orthodox decision theory15 guarantee that nothing relevant to one’s rational choice is left indeterminate. Second, orthodox decision theory takes the rationality of actions that require the exercise of agency through time to be determined by the rationality of each momentary choice; the extended action is rational if and only if each momentary decision maximizes utility. This is not to say that, on this view, our extended plans and ends play no role in our decision, but their role is exhausted by their contribution to the expected utility of each of the momentary choices that I make. As Paul Weirich puts it,
A wide variety of acts, such as swimming the English Channel, take more than a moment to complete . . . How does the principle of optimization address such acts? It does not compare the extended acts to alternative extended acts. Instead, it recommends an extended act only if each step in its execution is an optimal momentary act. Swimming the English Channel is recommended only if starting, continuing moment by moment, and finishing are all optimal. (Weirich (2004), p. 18)16
13 And for other theories as well, especially those that modify in some way or another orthodox decision theory. But focusing on orthodox decision theory has the advantage of letting us engage with a very clear and simple model.
14 More on these commitments of orthodox decision theory below.
15 For example, the axiom of continuity and the axiom of completeness, axioms that, roughly, require that the agent has an attitude of preference or indifference for any two possible options. More on these axioms below.
16 For an interesting, similar idea in a different context, see Portmore’s defence of maximalism in Portmore (2019) (especially chapter 4). For Portmore the evaluation of the extended action is prior to the evaluation of the momentary actions, but a similar biconditional still holds.
However, these two points express assumptions that are far from obvious. The assumption that rationality requires a complete ordering finds few adherents. Even philosophers who are sympathetic to normative versions of decision theory tend to reject the claim that ideal rational agents must satisfy the axioms, such as the axiom of continuity and the axiom of completeness.17 Indeed why should fully determinate preference ordering, or fully determinate ends, be a condition of instrumentally rational agency? As Cindy Lauper has wisely pointed out, sometimes rational agents (or at least some rational agents) might “just wanna have fun”, and it is not clear that this end requires a determinate conception of what counts as fun. The assumption that principles of rationality apply primarily or fundamentally to momentary actions has been mostly taken for granted in the philosophical literature,18 but it is far from intuitively obvious. If my ends extend through time (“having swum the English Channel”) and are pursued through means that extend through time (“swimming through the English Channel”), why would the rationality of these pursuits reduce to momentby-moment decisions? Of course, these rhetorical questions are not a proper argument against these assumptions and quite a great deal of the book will try to present a compelling picture of instrumental rationality that does not depend on these assumptions. But it might be worth keeping in mind that instrumental rationality should, by its very nature, make as few substantive assumptions as possible. It leaves such assumptions to a theory of substantive rationality (if there is such a thing) and only looks at what a rational agent must pursue given what the agent is either already inclined to pursue, or committed to pursue, or simply pursuing.
1.2 What is a Theory of Instrumental Rationality?
The awkward disjunction at the end of the last sentence in the previous section is a symptom of a more general difficulty in understanding the subject matter of a theory of instrumental rationality. In a traditional conception, instrumental rationality was understood in terms of the principle of instrumental reasoning, a principle which connects ends with means. Having ends and taking means can be understood as certain kinds of mental states
17 See, for instance, Joyce (1999) and Weirich (2004) itself.
18 In fact, even Weirich’s explicit acceptance of the assumption is relatively rare. Most of the time, the assumption is not even mentioned.
(intentions), activities (willings), or even kinds of actions (intentional actions). Such conceptions, however, might seem quaint and antiquated to our postVon Neumann and Morgenstern gaze. A conception of instrumental rationality as utility maximization is one in which one kind of mental state (preference) relates to another kind of mental state (choice); such a conception underlies the remarkably fruitful orthodox decision theory of Von Neumann and Morgenstern (2007) and Savage (1972).
Traditional and contemporary theories not only have different conceptions of the entities that are relevant to a theory of instrumental rationality but, under certain understandings of the former, they might disagree even about the ontological type of the entities in question. Although some philosophers view decision theory as a theory about how the agent effectively pursue her ends, the notion of an “end” does not appear in the theory; neither utility nor preferences map unobtrusively into the notion of an end. It would thus be helpful if there were some general way to characterize instrumental rationality that is neutral between these various versions. What I propose below is a rough way of doing this. I’ll not try to say much in defence of this framework, as I only intend it to be a useful way of capturing a wide array of different views under the same rubric of a theory of instrumental rationality. I doubt that any of the arguments of this book will depend on this particular framework.19
Let us start with an ideally (instrumentally)20 rational agent. Our ideal rational agent must be finite. After all, God does not take means to ends or need to make choices in light of her preferences; God’s representation of the ideal world directly causes it. Our ideal rational agent will have some representational attitudes21 that will cause or explain22 her actions. Among these attitudes that can cause the agent to act, there will be kinds of representational
19 Of course, someone might disagree that orthodox decision theory and the principle of instrumental reasoning even share a subject matter. Hampton (1998) makes a suggestion along these general lines. Even this position is compatible with the aims of the book, however, since the arguments of the book ultimately don’t depend on the assumption that they share a subject matter. For a view on the relation between the axioms of decision theory and the means–end principle that clearly implies that they do share a subject matter, see Nozick (1994). At any rate, not only do I believe that it is implausible to reject this assumption, I also think that it makes for much easier presentation to accept it.
20 I will often drop the qualifier “instrumental” (and related qualifiers) when the context makes it clear that we are talking about instrumental rationality.
21 I’ll try to avoid talking about “propositional” attitudes as I do not want to take a stand on whether the attitudes in question take a proposition as their content. But they’ll have some representational content.
22 I am not committing myself to causalism here (the “cause” in question need not be efficient cause).
attitudes that are such that they always motivate an agent insofar as she is instrumentally rational. In other words, a rational agent will make the content of these attitudes actual if it is within her power to make it actual and nothing interferes (possible interferences include adverse conditions, ignorance and error, or simply the existence of other attitudes of this kind with incompatible contents). Let us call these attitudes “motivationally efficacious representational attitudes”, or just “efficacious attitudes”. An intention, for instance, would be a candidate for this kind of attitude. A rational agent would make the content of her intention actual if nothing interferes, and so forth. Among motivationally efficacious attitudes, there will be some that are, from the point of view of instrumental rationality, given or basic (I’ll call them just “given attitudes” or “basic given attitudes”). The given attitudes provide, in some way, the standard of success for the theory and are not subject to direct evaluation in the theory of instrumental rationality. So if intending something (as an end) is the given attitude for a certain theory, then, first, the correct principle of rationality must be somehow connected to (the expectation of) making the end actual. Second, in itself, the fact that an agent intends an end is not a subject of evaluation for the instrumental theory of rationality. And as long as the object of intention is a coherent object to be made actual (it is not, for instance, a logically impossible object), the theory of instrumental rationality does not evaluate the intention as either rational or irrational.
Suppose desires are our given attitudes. In this case, the content of desire would provide a basic standard of success for the theory of rationality; when everything goes as well as possible,23 the rational agent satisfies her desires. One can now have many reasons why desires provide this standard of correctness; one might have a Humean view of the relation between passion and reason, one might think that desires when working properly disclose what is valuable to us, and so forth. Other candidates for taking the role of given attitudes are preferences, (evaluative or practical) judgments, intentions, and intentional actions.24 It is worth noting a few things: nothing I said commits us to accepting that a theory of instrumental rationality can allow for only one type of given attitude. A theory might allow multiple attitudes of this kind: arguably, philosophers who defend the importance of
23 Of course, once we allow for the possibility of risk, rational agency cannot guarantee that things go as well as possible. In such cases, moving from the content of the given attitudes to what counts as success is significantly more complicated. But I’ll ignore this complication except when I explicitly discuss cases of risk.
24 This is not meant to be an exhaustive list.
future-directed intentions in a theory of instrumental rationality will accept that both preferences and intentions are given attitudes. On the other hand, we also do not rule out in advance the possibility that only members of a subset of attitudes count as given attitudes. So one might allow only informed preferences into our set of given attitudes, or only intentions for permissible ends. Such moves would rule that an agent could not count as instrumentally rational when responding to, respectively, misinformed preferences or evil intentions. This qualification is particularly important if one takes the principles of instrumental rationality to be “narrow scope”,25 in which case failing to limit their application in this way might have counterintuitive consequences. But it also makes it possible to allow for various principles proposed even by “myth theorists”26 to constitute a theory of instrumental rationality. For we could allow that the given attitudes are “knowledge (justified beliefs) about reasons”, as long as the “facilitative principles” proposed by philosophers like Raz (2005) and Kolodny (2018) imply similar principles governing how an agent correctly respond to facts that stand in instrumental relations to the reasons they have. An instrumentally rational agent, on such a view, would be one who responds to the reasons they know (justifiably believe) in accordance with (or based on) the Facilitative Principle.27
Finally, one might object to the inclusion of intentional action28 as a possible given attitude; after all, intentional actions are not usually regarded as attitudes. Since this is the view I favour, I should say a few words about it. It is certainly correct that intentional action is an event or process in the external world.29 However, what constitutes it as an intentional action is in
25 For discussion of the scope of the principle of instrumental reasoning, see, among many others, Greenspan (1975), Broome (1999).
26 Kolodny (2005, 2007), Raz (2005).
27 A similar formulation could be given by using Lord (2018)’s conception of a possessed reason.
28 I’ll often just refer to “action” as opposed to “intentional action”. Hyman (2015) warns that philosophers run together the distinction between action and non-action, voluntary and involuntary, and intentional and non-intentional, but I’m not sure that this is entirely fair (for an earlier careful discussion with some of these issues, see Anscombe (2005)). At any rate, my usage will be restricted to cases in which the action (or refraining, or omission) is a manifestation of our will, considered as a rational faculty (although I agree with Kant that the “will is nothing but practical reason” (Kant (1996), Ak. 6: 412), alignment to the party line is unnecessary for the argument of the book). In the language we are about to develop, what matters for our purposes is that the relevant attitude is the outcome of the exercise of (or failure to exercise) the relevant rational powers relative to a given attitude.
29 I am not settling the issue about the correct ontological category of actions, though a view of actions as processes is more congenial to the view I defend here. See Steward (2012).
part, on any view, a representation of the agent’s. On Davidson’s view (Davidson 1980a), these are beliefs and desires that predate the action, and on some Davidsonian views, intentional actions are constituted by certain mental states that causally sustain the action.30 On an Anscombean view,31 practical knowledge, understood as the agent’s knowledge of what she is doing and why, is constitutive of intentional action. On a more classical view, animal action in general is the effect of the action’s representation in the faculty of desire,32 and at least in some of these views, rational action in particular is determined by an attitude that represents the action as good.33 Although in the Anscombean view, intentional action parallels most clearly the structure of other propositional (or contentful)34 attitudes, all these views accept that intentional actions share with mental attitudes an important feature for our purposes: they create opaque contexts that track the agent’s point of view; what the agent does intentionally depends on the agent’s perspective on the action. I cannot be intentionally ϕ-ing if the fact that I am ϕ-ing is in no way available to my mind.35 This is just Anscombe’s (later endorsed by Davidson) classic observation that actions are intentional only under a description36 (and that the correct description depends on the agent’s point of view).
A theory of instrumental rationality also needs to identify what counts as an exercise of the relevant practical capacity, or more generally it needs to specify the nature of the practical capacity it governs. If God desires light, that very representation will assure that there is light; even the command “let there be light” seems an unnecessary adornment: at best, it is a kind gesture to less powerful beings like us who would otherwise fail to recognize light as the effect of divine judgment. But finitely rational beings cannot always bring things about merely by desiring, intending, representing as good, and so forth. Finite beings must (often) exercise some active power to make the content of their given attitudes actual,37 and the business of a
30 Setiya (2007, 2011).
31 Anscombe (2000).
32 “The faculty of desire is a being’s faculty to be by means of its representations the cause of the reality of the objects of these representations” (Kant (1996), 5:9n).
33 “It is an old formula of the schools, nihil appetimus, nisi sub ratione boni . . . [This] proposition if it is rendered: we will nothing under the direction of reason insofar as we hold it to be good is indubitably certain” (Kant (1996), 5:59–60).
34 This clause allows for views in which the content of the attitude is not a proposition.
35 This is much weaker than the “belief” condition on intentional action; namely, that I can only ϕ intentionally if I believe I am ϕ-ing intentionally.
36 Anscombe (2000) and Davidson (1980a).
37 On some conceptions of the given attitude they will themselves to be exercises of some active power (this will actually be the view proposed here). But even such views will need to
