Why Future-Bias Isn’t Rationally Evaluable

Future-bias is preferring some lesser future good to a greater past good because it is in the future, or preferring some greater past pain to some lesser future pain because it is in the past. Most of us think that this bias is rational. I argue that no agents have futurebiased preferences that are rationally evaluable—that is, evaluable as rational or irrational. Given certain plausible assumptions about rational evaluability, either we must find a new conception of future-bias that avoids the difficulties I raise, or we must conclude that future-biased preferences are not subject to rational evaluation. Future-bias is typically thought to be rational, often obviously rational.1 Parfit (1984) offers this now familiar example. I am in some hospital to have some kind of surgery. Since this is completely safe, and always successful, I have no fears about the effects. The surgery may be brief, or it may instead take a long time. Because I have to cooperate with the surgeon, I cannot have anesthetics. I have had this surgery once before, and I can remember how painful it is. Under a new policy, because the operation is so painful, patients are now afterwards made to forget it. Some drug removes their memories of the last few hours. I have just woken up. I cannot remember going to sleep. I ask my nurse if it has been decided when my operation is to be, and how long it must take. She says that she knows the facts about both me and another patient, but that she cannot remember which facts apply to whom. She can tell me only that the following is true. I may be the patient who had his operation yesterday. In that case, my operation was the longest ever performed, lasting ten hours. I may instead be the patient who is to have a short operation later today. It is either true that I did suffer for ten hours, or true that I shall suffer for one hour. 1 Others have called this ‘bias against the past’ or ‘past discounting.’ Res Philosophica, Vol. 98, No. 4, October 2021, pp. 573–596 https://doi.org/10.11612/resphil.2143 © 2021 Callie K. Phillips • © 2021 Res Philosophica 574 Callie K. Phillips I ask the nurse to find out which is true. While she is away, it is clear to me which I prefer to be true. If I learn that the first is true, I shall be greatly relieved. (165–166) It seems rational to prefer to be in the first situation in such cases despite the fact that this preference involves preferring to be in a situation that has a greater total amount of pain for you in the overall course of your life. In fact, many philosophers take it to be obvious that this preference is rational and without need of defense. Prior (1959), Craig (1999), Zimmerman (2005), Heathwood (2008), Hare (2007, 2009), and Kauppinen (2018) all claim that this bias is at least rationally permissible, if not rationally required. However, a handful of philosophers—including Moller (2002), Brink (2011), Sullivan and Greene (2015), Dougherty (2015)—raise difficulties for defending the rationality of this kind of bias. In this article I present an argument that would mean that all these authors have got it wrong. Future-biased preferences are not even rationally evaluable—that is, evaluable as rational or irrational. In determining whether an agent is future-biased, I will make use of the definitions offered by Sullivan and Greene (2015). An agent S is [future-biased] with respect to pleasure iff for two exclusive experiences, E1 and E2, where E1 is at least as pleasurable as E2, S prefers E2 because it is a present or future pleasure rather than a past one. . . . An agent S is [future-biased] with respect to pain iff for two exclusive experiences, E1 and E2, where E2 is at most as painful as E1, S prefers E1 because it is a past pain rather than a present or future one. (949) I will call agents “future-biased” if they are future-biased with respect to pleasure, pain, or both. I will say that a preference is a future-biased preference if it is a preference that meets the description on the right-hand side of one those biconditionals. In this article I argue that we must reject a plausible assumption about rational evaluability, find a new conception of future-bias that avoids the difficulties I raise, or conclude that future-biased preferences are not subject to rational evaluation. The conclusion that there are no rationally evaluable future-biased preferences is surprising and counterintuitive, but it has a substantial theoretical advantage that makes this neglected view worthy of our attention. One who is willing to accept the conclusion that future-biased preferences are not rationally evaluable will be freed from the burden to explain the widely assumed irrationality of another very common sort of time-bias—namely, near-bias. How to discharge this burden has eluded philosophers since Parfit first made the case for the tension between these biases in Reasons and Persons (1984). Near-bias, roughly, is preferring less pleasurable experiences because they are nearer in temporal proximity, or, conversely, Why Future-Bias Isn’t Rationally Evaluable 575 preferring more painful experiences to less painful ones because those more painful experiences are further in temporal proximity. A popular putative example of near-bias is found in the well-known Stanford marshmallow experiment. This study sought to measure the selfcontrol of children through an experiment in which some children were told they could eat one marshmallow now or wait fifteen minutes and receive two marshmallows. The kids who didn’t wait the fifteen minutes to receive the two marshmallows are said to be near-biased—they prefer a lesser good now to a greater good in the future.2 The claim that near-bias is irrational seems to hardly warrant defense.3 But Parfit (1984) argues convincingly that it is very difficult to defensibly claim both that near-bias is irrational and that future-bias is rational. In brief, the argument is that the way to argue that near-bias is irrational is to claim that a mere difference in when something occurs does not affect its value. But if we make this claim we should also think that future-bias is irrational. The fact that some experience is past shouldn’t affect its value for us. Parfit considers possible replies to this argument, but none of them seem to withstand scrutiny. For example, you might think that future-bias is importantly different from near-bias because we can’t change the past. If we can’t change the past, it is reasonable for us to be less concerned about it.4 However, we still seem to exhibit future-bias even if the future experience in question cannot be changed. If it were certain that you were going to be tortured later this afternoon, you would still be more concerned about this future painful experience than you would be if it were already in the past. Parfit considers and dismisses other possible defenses of an asymmetry in the rationality of near-bias and future-bias. In my view, these arguments have yet to be challenged successfully. Thus, it’s important to find a way to explain or 2 Mischel et al. 1972. It is worth noting that there is now doubt that the children in the study were near-biased in the sense that interests us (i.e., there’s doubt that they had preferred the temporally nearer marshmallow eating experience simply because it was temporally nearer). Delaying gratification, waiting to receive two marshmallows had a lot to do with economic and social disadvantages (Watts et al. 2018). It seems children who didn’t wait, often from resource poor environments, doubted that there would be marshmallows in the future, or at least assigned it lower probability. If a later outcome is lower probability than the nearer outcome, it will often be rational to choose the nearer. This highlights one difficulty with empirical investigation of near-bias and future-bias. It is difficult to create conditions that would clearly demonstrate a preference for one experience over the other based simply on the temporal location of those events (near or far, past or future) since it is difficult to create conditions where the likelihood of those experiences is the same. 3 For philosophers and the casual observer at least. The irrationality of near-bias is more controversial among economists and psychologists who are more likely to hold that only structural constraints on preferences are relevant for whether preferences are rational. 4 This strategy for explaining how future-bias is rational while near-bias is irrational has been pursued by Suhler and Callender (2012) and Dyke and Maclaurin (2002). Latham et al. (2021) investigate empirically the reasons for future-biased preferences and find that the belief that the past is practically irrelevant plays a role, but also find that even if agents supposed that they could causally affect the past, they still exhibit future-bias. 576 Callie K. Phillips avoid the asymmetry of the rationality of these time-biases. If one accepts my argument that future-biased preferences are not rationally evaluable, there is no asymmetry to explain; this major obstacle for defending the irrationality of near-bias falls to the wayside. In Section 6, I discuss this theoretical advantage and the implications of my argument in connection to near-bias in more depth. The argument of this article shows that the position that there are no rationally evaluable future-biased preferences can be motivated and is more attractive than it initially appears. Even if the best response to my argument is to reject one of the premises, we’ll learn that we must eschew plausible claims about preferences and rational evaluability—a lesson that has gone unappreciated. 1 Representation and Rational Criticism Our definitions of future-bias say that being future-biased is a matter of having certain preferences. So if we criticize the rationality of future-bias, we criticize having certain preferences. Whether it makes sense to criticize particular preference