Philosophical Studies

, Volume 156, Issue 3, pp 311–319

Lucky agents, big and little: should size really matter?



DOI: 10.1007/s11098-010-9595-z

Cite this article as:
Blumenfeld, D. Philos Stud (2011) 156: 311. doi:10.1007/s11098-010-9595-z


This essay critically examines Alfred R. Mele’s attempt to solve a problem for libertarianism that he calls the problem of present luck. Many have thought that the traditional libertarian belief in basically free acts (where the latter are any free A-ings that occur at times at which the past up to that time and the laws of nature are consistent with the agent’s not A-ing at that time) entail that the acts are due to luck at the time of the act (present luck) rather than to the kind of agent control required for genuinely free, morally responsible action. While libertarians frequently have tried to rebut the claim that basically free acts are due to present luck, Mele argues for the daring thesis that they should embrace present luck rather than try to explain it away. His strategy is to argue that the assumption of present luck in the decisions of very young children (or “little agents”) does not preclude us from attributing to them a small amount of moral responsibility and that this makes it possible to conceive of moral development as a gradual process in which as the frequency of the indeterministically caused free actions increases, the agents take on greater and greater moral responsibility. In this paper I give several possible reconstructions of Mele’s argument and analyze in detail why none of them succeeds.


Daring soft libertarianismPresent luckLittle agentsFree willMoral responsibility

A persistent worry about libertarianism is how, if indeterminism is required for free will, there can be a sufficiently tight connection between an agent and her actions to allow for moral responsibility. The worry comes in two varieties, an older one, which has been widely discussed, and a newer one that is still under debate.1 I focus here on the newer one, in particular on its deployment by Alfred R. Mele (2006) and especially his attempt to allay a worry about “present luck” (luck at the time of action) for a view he calls “daring soft libertarianism.” (DSL = daring soft libertarian).2 Although few have pressed the issue as relentlessly as Mele, his distinctive resolution of it has yet to receive the attention it deserves. In this essay, I describe and criticize Mele’s treatment of the issue.


The problem of present luck arises for all libertarians who think, as they traditionally have, that the existence of free action depends on there being basically free actions, which are “any free A-ings that occur at times at which the past (up to those times) and the laws of nature are consistent with the agent’s not A-ing then.” As to free A-ings themselves, Mele says that libertarians can either claim that an agent freely A-ed at t only if he could have done otherwise at t or that an agent who could not have done otherwise at t may still freely A at t, provided that he earlier performed some relevant free action at a time when he could have done otherwise (2006, p. 6). In other words, libertarians can hold that basically free actions are the only free actions there are or that basically free acts that are appropriately related to a person’s subsequent actions confer freedom on the latter, even though the agent could not have done otherwise at the time. Parallel options are open for morally responsible action. But as long as libertarians believe in basically free acts, they will have to deal with the problem of present luck, which Mele sets out as follows.

Suppose Joe performs a basically free act of deciding at t to A.3 By hypothesis, there is another possible world with exactly the same past and laws of nature in which at t he does not decide to A. Imagine that at this other world Joe freely decides at t not to A.4 Since everything in the two worlds is the same up to t, including all of Joe’s powers, states of mind, and the like, there is nothing in Joe to account for his deciding to A in the one world and not to A in the other. In that case, the cross-world difference in his decisions is just a matter of luck (good luck if A is a good decision, bad luck if it is a bad one) and Joe does not have the control required for a free decision (2006, pp. 7–9).

Some may point out that this indeterminism is compatible with there being objective probabilities about what Joe will decide, and if it is more probable that he will decide to A than not to A, his deciding to A is not just a matter of luck. True, says Mele, but Joe’s decision is partly a matter of luck and the difference between the world in which he decides to A and the one in which he decides not to A is just a matter of luck (2006, pp. 8–9, 114, 132). So the problem remains.

One can’t solve it, moreover, by arguing that Joe’s decision at t will be free and responsible if the probabilities at t result from Joe’s earlier basically free decisions, as the worry about present luck affects his earlier decisions too (2006, pp. 61, 74). Ultimately, then, whoever thinks freedom and responsibility require basically free actions must explain how this is consistent with present luck.

It is a measure of the force of this argument that so many contemporary libertarians have been impressed by it. Peter van Inwagen, for example, who, in his classic study, An Essay on Free Will, was dismissive of criticisms based on chance has reversed course and given an argument based on a worry similar to Mele’s (1983, pp. 128–129, 2002, pp. 171–172), declaring that, until it is resolved, free will remains a mystery (2002, p. 171).5 Robert Kane (1989, 1999), Randolph Clarke (2002, 2003, 2005) and Timothy O’Connor (2000), on the other hand, have taken great pains to try to solve the problem and have offered inventive responses to it. But ingenious as these attempts are, they remain controversial; and Mele, who argues that none of them works (2006 and passim), sets out to finish the job.6


What is his solution? While some libertarians have tried to explain away present luck, DSLs “embrace” it, stressing that present luck is implied by the kind of power they value and arguing that it is no bar to freedom or responsibility. DSLs also hold that basically free acts7 that are suitably related to the agent’s later actions confer freedom and responsibility on them even if the agent could not do otherwise at the time. In virtue of her earlier basically free acts, an agent may likewise be responsible for the practical probabilities that govern her later actions (2006, pp. 113–116, 121–123). But, by itself, this just pushes the issue to an earlier time and threatens to generate a vicious regress. The challenge is to stop the regress.

Mele’s plan is to show that small children, or “little agents,” can have a bit of responsibility for their basically free decisions which is not conferred from the past. This will stop the regress and allow DSLs to explain the acquisition of greater moral responsibility as a gradual process in which, as the frequency and range of little agents’ basically free actions increases, they take on increased responsibility for their actions and the practical probabilities thereof (2006, p. 132).

To support this, Mele cites several key facts about little agents. One is the trivial nature of their decisions, another is the limited responsibility we ascribe to little children. He says, for example, “the good and bad deeds of young children are relatively trivial in themselves” and “not that big a deal.” Thus, little agents “are not nearly as morally responsible for any of their deeds as some adults are for theirs” and “only a relatively modest degree of moral responsibility is at issue” (2006, pp. 130–132). He also mentions two other factors: that young children are “not as well equipped for impulse control” as older children or adults normally are; and that young children have “a less developed capacity for anticipating and understanding the effects of their actions” (2006, p. 129).

What is there to persuade us that these traits are consistent with little agents having the required moral responsibility? In response, Mele asks us to consider the first time a normal 4-year-old, Tony, makes a decision about whether to snatch a toy from his younger sister. Suppose that, frustrating though it is, Tony decides not to snatch the toy. Surely, Mele says, Tony’s father would regard him as deserving of at least a little credit for his action just as he would have regarded him as deserving of a little blame had Tony snatched the toy. Now, since we are assuming that Tony’s decision is basically free, it follows that there is another possible world with the identical past and laws of nature at which he decides to take the toy and that the difference between the two worlds is just a matter of luck. But, Mele asks, would this lead us to conclude that Tony has no moral responsibility for his decision? Not even a tiny bit that would make him deserving, say, of a pat on the head or a smile and a few words of praise? On the contrary: when we consider how trivial their actions are, how immature their capacities are and what a modest a degree of responsibility we ascribe to little agents.

we should be struck by the implausibility of stringent standards for deserved moral praise and blame of young children—including standards [requiring] the absence of present luck (2006, pp. 131–132).


An initial problem is that if we ask what undercuts responsibility for Joe (the adult agent Mele used to set up the problem), it seems to be only that the difference between the world at which he decides to A and the one at which he decides not to A is just a matter of luck. In fact, we aren’t told anything else: only that Joe decides to A and that this involves present luck. More isn’t required, it seems, because present luck alone negates moral responsibility. But then since present luck is at hand in both cases, little Tony shouldn’t be responsible if big Joe isn’t. Why should size matter?

For DSLs, the answer lies in the factors that differentiate little agents from big ones, and I think that LA is a plausible way of capturing their reasoning.

Little Agents Argument (LA)
  1. (1)

    If an agent makes a basically free decision and (a) it is relatively trivial; (b) it is much harder for the agent to exercise impulse control over it than it normally would be for an or older child or adult; and (c) the agent is far less able to appreciate its consequences than an older child or adult normally would be, then the agent has a small degree moral responsibility for it even though it involves present luck.

  2. (2)

    Little agents make basically free decisions and (a)–(c) hold of those decisions.

  3. (3)

    Hence: Little agents have a small degree of moral responsibility for their basically free decisions even though they involve present luck.

We should bear in mind that while DSLs are committed to LA, Mele is not. He is agnostic about whether there are any basically free decisions (footnote 6) and so he does not accept LA 2. His aim is only to show that assuming that there are basically free decisions and that the decisions of little agents are among them, the present luck they entail does not preclude moral responsibility. Since my interest here is only in whether Mele proves this claim, in assessing LA I will grant for the sake of argument that these DSL beliefs are correct. By doing so, I will be discussing just the part of the argument Mele accepts. In light of these things, how should we judge LA?

To begin with, I think LA1 is implausible since I don’t see why (a)–(c) should block the effects of present luck. In fact, on assumptions I discuss below, cases are possible in which the antecedent of LA1 is true and its consequent false. But before showing this, let me explain my strategy and defend its assumptions.

For one thing, I will assume that adult basically free decisions that satisfy (a)–(c) are possible, and I will describe a case of this sort. This presents no problem for a DSL, nor does anything in DSL theory entail that there could not be such a case. DSLs believe that there are adult basically free decisions; and even though adult decisions normally do not satisfy conditions (a)–(c), in fact they sometimes do. So I can see no reason why adult basically free decisions could not sometimes satisfy conditions (a)–(c) or why a DSL should think otherwise. In order to make my case parallel to that of little agents, I will also assume that the adult decision in it does not involve responsibility conferred from the past. This assumption is legitimate too: DSLs do not maintain that adult basically free decisions always involve conferred responsibility.8 More important: even if DSLs did hold this, since my strategy is to challenge LA and the idea of conferred responsibility that rests on it, an appeal to conferred responsibility at this point would be question-begging. My aim is to show that when we consider a case like the one I describe, we will judge that the agent is not responsible, provided of course we think, as Mele does, that present luck is a prima facie problem for more consequential adult decisions. This will establish that LA1 is false by giving an example in which its antecedent is true and its consequent false.

To execute this strategy, let’s go at premise 1 bit by bit. That is, let’s first craft a case of a trivial adult basically free decision for which the agent is not morally responsible. This will show that part (a) of the antecedent of 1 can be true and the consequent false. I will then alter the case so that (b) and (c) are also true but the consequent remains false. Here’s the initial example.

Cathy asks Felix if he will feed her cat, Indulgence, while she is away for a week. As Cathy explains, Indulgence will eat only two brands of cat food, Cat Caviar, which she will tolerate if necessary, and Pussy Pleaser, which she is crazy about, and which Cathy now feeds her exclusively. If Felix is willing to do the favor, he must promise to feed Indulgence only Pussy Pleaser. He promises and for 6 days performs as agreed. But when he arrives on the final day, he finds that Cathy has only Cat Caviar on hand. Since he doesn’t want to go out to get some Pleaser, Felix decides: “Let her eat Caviar!” and he feeds Indulgence her less favored food.

If we make no special metaphysical assumptions about this case and just judge it from an everyday perspective, I’d say Felix has a small amount of responsibility for his decision. But that’s not the correct perspective: we are assuming that Felix’s decision involves present luck and that his earlier decisions have not conferred responsibility on him for this one. If we keep this in mind and if we assume with Mele that (barring an independent proof to the contrary) the present luck in Joe’s decision would exempt him from responsibility, how can we claim (barring a similar proof) that Felix is morally responsible for his decision, even if only to a small degree?

So conjunct (a) of premise 1 can be true and the consequent of 1 false. But this doesn’t falsify 1 since its antecedent also includes (b) and (c). Could it be that if (b) and (c) were also true, then in no instance would the consequent of 1 be false? I do not see how this could possibly be so. If a basically free agent, whether big or little, is not responsible for a decision, how could adding (b) (that it was especially hard for him to exercise impulse control) and (c) (that he was less able to appreciate the consequences of his decision than an adult or older child normally would be) make him more responsible than if (b) and (c) had not obtained? These factors tend to diminish control and thereby reduce responsibility.

Some may object that LA incorrectly depicts the DSL’s argument. As it stands, LA1 covers any basically free decision of which (a)–(c) are true, whereas it may seem that Mele restricts LA to a little agent’s first such decision. Since we don’t make our first basically free decisions as adults, this would block my point.

Although Mele does ask us to consider Tony’s first basically free decision, it is not clear to me that he intended to make this restriction. Perhaps he meant to make the general claim expressed in 1 and was just thinking that since (a)–(c) are true of first decisions, this would serve his purposes. But, in any case, I don’t think the restriction is defensible, since it is hard to see what justification it could have. Why should a decision’s being the agent’s first one affect whether it is made freely and responsibly, which have nothing per se to do with the temporal order or time of life in which a decision is made? Without a justification the restriction is unwarranted.


Let’s turn to premise 2. LA2 says that little agents make basically free decisions and (a)–(c) hold of those decisions, where (a) specifies that the decisions are relatively trivial. But while it may be that little agents’ basically free decisions are usually trivial, they aren’t always so. At a tender age, some twin boys in my family took to biting one another so savagely and in such tender places that their parents had to monitor them constantly for fear of their doing each other irreparable harm. In another case, a friend of mine repeatedly had to restrain her young daughter (about age 3) from climbing into the crib with her infant brother and tightly hugging his blanket around his head in order to ‘love’ him (to death, no doubt). Such cases are not uncommon. So, as it stands, 2 is false because little agents’ basically free decisions are not always trivial.

The remedy, of course, may seem obvious. All the DSL needs (and perhaps all Mele intended) is for 2 to say that little agents’ basically free decisions sometimes satisfy (a)–(c). This won’t save LA, since 1 is false, but at least it will block the criticism of 2.

This raises the question, however, of what the DSL is to say about little agents whose basically free decisions are not trivial but of which (b) and (c) are true. There are a limited number of options here, only one of which holds up. It clearly won’t do to say that if there are early basically free decisions of which (b) and (c) are true, they merit some responsibility if they are trivial but none if they are significant. There is nothing to recommend this curious proposal.

A more plausible but still unsatisfactory option is to hold that the non-trivial decisions in question generate a moderate degree of moral responsibility. One might think that since the decisions are non-trivial, they must generate more than just a little responsibility but, because (b) and (c) hold, the amount they generate cannot be very great either. So it’s somewhere in-between. The problem is that we are discussing little children — mere babes, whose capacity for impulse control, as Mele has stressed, is extremely limited and whose ability to appreciate the consequences of their decisions, is rudimentary. This makes it hard to support the claim that such agents have more that a very small amount of moral responsibility for their decisions, trivial or otherwise.9

The best option is to say that, like their trivial ones, little agents’ basically free, non-trivial decisions generate only a small amount of responsibility. Even when young children’s decisions are non-trivial, conditions (b) and (c) will keep their moral responsibility very limited.

But in that case, even though DSLs can skirt the criticism of 2, our discussion of it was instructive, since showing that little agents’ basically free, non-trivial decisions also produce a small amount of responsibility reveals that the triviality condition is not essential to LA. A version without it would begin: If an agent makes a basically free decision and (a) it is non-trivial and (b) and (c) hold of it, then the agent has a small degree of responsibility for it even though it involves present luck.

An argument based on this premise fails too, however. Suppose Felix promises to feed Cathy’s sick grandmother for a week,10 but unexpected side-effects of a new medicine he is taking cause (b) and (c) to obtain, though without determining his decisions. On day 7, Felix decides he’d rather go fishing than feed Granny. Since his decision is basically free, there is a possible world with the same laws and past at which he decides to keep his promise and the difference between that world and the actual one is just a matter of Felix’s bad luck. So Felix is not morally responsible for his decision.

I conclude that the DSL’s embrace of present luck is unsuccessful: where present luck is at hand, agent size doesn’t matter.


The older worry is: If freedom requires indeterminism, free acts are either uncaused or indeterministically caused. If they are uncaused, they are obviously due to chance. If they are indeterministically caused (e.g., by the agent’s desires and beliefs), they are still due to chance since, given indeterminism, with identical factors present the agent might have acted otherwise. But free acts can’t be due to chance.


Mele calls DSLism soft because it rejects the hard-line libertarian view that freedom entails indeterminism. It holds that a legitimate species of freedom and responsibility is compatible with determinism but maintains that the falsity of determinism is required for a more desirable species, which a person may rationally prefer to any provided by compatibilism (2006, p. 103 n. 18). He regards the position as daring because it both embodies the traditional libertarian thesis that moral responsibility presupposes that there are basically free actions (as defined in section 1 of this essay) and also embraces the idea that such decisions entail present luck. DSLism contrasts with an alternative theory of his, namely, modest soft libertarianism, which rejects the traditional libertarian idea that moral responsibility presupposes basically free actions. Also see note 6.


Mele’s theory is a form of event-causal libertarianism that focuses on decisions which, for him, are momentary mental acts (2006, pp. 15, 57).


Mele first raises the issue using a case in which the act is basically free because the agent freely A’s at t and there is a possible world in which at t he freely does not-A. In chapter 3, he extends the argument to other cases of basically free actions (e.g., ones where the act is basically free because there is a possible world in which at t the agent does something other than A, albeit unfreely) and eventually to “basically*” free actions (see note 7). But since these steps do not affect my critique, I discuss only the case at hand.


On the differences between van Inwagen’s and Mele’s versions of the problem, see Clarke (2005, pp. 413–414).


NB. Despite his attempt to solve the problem, Mele is not a libertarian. Although he believes in freedom and responsibility, he is agnostic about whether their requirements are compatibilist or incompatibilist and whether there are any basically free acts. So rather than try to establish compatibilism or incompatibilism, he instead develops the best versions he can of each but is non-committal about which is correct (1995). Ultimately, he constructs two incompatibilist theories, the first of which is a modest soft libertarian view that locates indeterminism not at the time of decision but earlier, in indeterministically caused factors such as beliefs and desires that happen to occur to the deliberating agent. The second is DSLism.


Actually, in developing his theory Mele shifts to the notion of basically* free actions (2006, p. 115), which is designed to accommodate the DSL’s belief that some Frankfurt-style cases are persuasive. But because this is irrelevant to my argument, we can safely pretend that the notion in play is basically free actions.


Referring to adult basically free (actually, basically*—see note 7) Mele says that “in the vast majority” of cases “agents have some responsibility for the relevant practical probabilities,” thus implying that there are exceptions to the general rule that such actions involve conferred responsibility (2006, p. 123).


I think, for example, that the 3-year-old who swaddled her infant brother’s head would have had at most a very small degree of moral responsibility even if she had actually killed him, since 3-year-olds have not yet developed a concept of death (Speece and Brent 1984, 1992; Hunter and Smith 2008).


But not with Pussy Pleaser!.



For helpful comments I am indebted to Paul Bassen, Robert Kane, Eddy Nahmias, David Widerker and an anonymous referee for this journal.

Copyright information

© Springer Science+Business Media B.V. 2010