Alfred Mele has developed the idea of autonomous agents, who among other things exercise a kind of self-control that is related to metaphysical freedom terms like "free will" and "free action." He has also developed a number of models for free will, most notably his 1995 two-stage model called "Modest Libertarianism. Others include Soft Compatibilism, Soft Libertarianism, and Daring Soft Libertarianism
Without committing himself to the idea that human autonomy is compatible with determinism or incompatible (the position of the libertarians), Mele provides arguments in support of autonomous agents for both positions. He is, as he says, "officially agnostic about the truth of compatibilism" and describes his position as "agnostic autonomism."
Mele's opponents are those who believe there are no free and morally responsible human beings. They are philosophers who deny both compatibilism and libertarianism - Richard Double and Ted Honderich, for example, "Impossibilists" like Galen Strawson, and "Illusionists" like Derk Pereboom and Saul Smilansky. He has debated the psychologist Daniel Wegner, whose position is that the conscious will is an illusion. Note that Randolph Clarke's "narrow incompatibilism" denies the compatibilism of free will and determinism, but accepts the compatibilism of moral responsibility and determinism. John Martin Fischer's "semicompatibilism" similarly accepts the compatibilism of moral responsibility, while remaining agnostic about free will and the truth of determinism. Clarke's and Fischer's morally responsible agents presumably would be Mele "autonomous agents."
Most libertarians, Mele thinks, both agent-causalists like Timothy O'Connor and event-causalists, like Robert Kane, might subscribe to his "autonomous agent" idea.
"My plan in Mele [Autonomous Agents] 1995 was to use the resources both of libertarianism and of compatibilism in defending agnostic autonomism and to do that partly by developing the best compatibilist and libertarian positions I could develop. Part of my strategy was to construct an account of an ideally self-controlled agent (where self-control is understood as the contrary of akrasia: [a Greek term, meaning] roughly, weakness of will), to argue that even such an agent may fall short of autonomy (or free agency), and to ask what may be added to ideal self-control to yield autonomy (or free agency). I offered two answers, one for compatibilists and another for libertarians. I then argued that a disjunctive thesis associated with both answers—agnostic autonomism - is more credible than [believing there are no free and moral human beings] NFM." (Introduction, Free Will and Luck 2006, p.5)
Mele in his 1995 book Autonomous Agents proposed a "Modest Libertarianism" for consideration by libertarians. He himself did not endorse the idea. But he is concerned about the proper place to locate the indeterminism. His soft libertarians locate it somewhere in the chain of events leading up to the formation of intentions, the evaluation of options, the decision and ultimate action. His "daring soft libertarians" move the indeterminism up into the "time of action," where indeterministic alternative possibilities for actions may (or may not) exist.
He made it clear, following Daniel Dennett's "Valerian" model in Brainstorms, 1978, that any indeterminism should come early in the overall process. He even describes the latter - decision - stage of the process as compatibilist (effectively determinist). This of course could only be adequate determinism. Mele proposes a "soft compatibilism" that sees some value for indeterminism in the early stages.
"These observations indicate that it might be worth exploring the possibility of combining a compatibilist conception of the later parts of a process issuing in full blown, deliberative, intentional action with an incompatibilist conception of the earlier parts. For example, it might be possible to gain "ultimate control" while preserving a considerable measure of nonultimate agential control by treating the process from proximal decisive better judgment through overt action in a compatibilist way and finding a theoretically useful place for indeterminacy in processes leading to proximal decisive better judgments. (Autonomous Agents, p.212) location(s) for internal indeterminism would do us the most good. (p.213)Recall that compatibilism does not include a commitment to determinism. The thesis is that determinism does not preclude autonomy. Treating the process from proximal decisive better judgment through overt action in a compatibilist way does not require treating it in a determinist way. Compatibilists may, in principle be willing to accept an account of causation that accommodates both deterministic and probabilistic instances, and they are not committed to holding that probabilistic causation in the process just mentioned precludes the freedom of its product. In the same vein, advocates of autonomy who seek a "theoretical useful place" for indeterminism in the springs of action need not insist that indeterminism does not appear at other places, as well, in internal processes issuing in autonomous action. Their claim on that matter may merely be that indeterminism at these other junctures is of no use to them. (p.213) External indeterminism, as I have already explained, does not give libertarians what they want. That leaves internal indeterminism. Assume, for the sake of argument, that human beings sometimes act autonomously, that acting autonomously requires "ultimate control," and that the latter requires internal indeterminism. Then, with a view to combining ultimate control with robust nonultimate control, we can ask what
Mele's plan in his book Free Will and Luck is to pay more attention to Frankfurt-style examples and to "agential luck."
A Problem about Luck for Libertarians (p.7) Agents' control is the yardstick by which the bearing of luck on their freedom and moral responsibility is measured. When luck (good or bad) is problematic, that is because it seems significantly to impede agents' control over themselves or to highlight important gaps or shortcomings in such control. It may seem that to the extent that it is causally open whether or not, for example, an agent intends in accordance with his considered judgment about what it is best to do, he lacks some control over what he intends, and it may be claimed that a positive deterministic connection between considered best judgment and intention would be more conducive to freedom and moral responsibility. This last claim will be regarded as a nonstarter by anyone who holds that freedom and moral responsibility require agential control and that determinism is incompatible with such control. Sometimes it is claimed that agents do not control anything at all if determinism is true. That claim is false. As soon as any agent...judges it best to A, objective probabilities for the various decisions open to the agent are set, and the probability of a decision to A is very high. Larger probabilities get a correspondingly larger segment of a tiny indeterministic neural roulette wheel in the agent's head than do smaller probabilities. A tiny neural ball bounces along the wheel; its landing in a particular segment is the agent's making the corresponding decision. When the ball lands in the segment for a decision to A, its doing so is not just a matter of luck. After all, the design is such that the probability of that happening is very high. But the ball's landing there is partly a matter of luck. All libertarians who hold that A's being a free action depends on its being the case that, at the time, the agent was able to do otherwise freely then should tell us what it could possibly be about an agent who freely A-ed at t in virtue of which it is true that, in another world with the same past and laws of nature, he freely does something else at t. Of course, they can say that the answer is "free will." But what they need to explain then is how free will, as they understand it, can be a feature of agents — or, more fully, how this can be so where free will, on their account of it, really does answer the question. To do this, of course, they must provide an account of free will — one that can be tested for adequacy in this connection.
Mele proposes his "modest libertarianism" to satisfy these needs. It includes a two-stage process that first generates random alternative possibilities, which is then followed by a determination stage. When he first mentioned his idea in 1995, Mele cited the similar "Valerian" example Daniel Dennett had proposed in 1978 as something libertarians should want.
Note that both Dennett and Mele are skeptical that any such process exists, but note that Mele's model does indeed satisfy most of the requirements for libertarian free will.
A Modest Libertarian Proposal (p.9) According to typical event-causal libertarian views, the proximate causes of free actions indeterministically cause them. This is a consequence of the typical event-causal libertarian ideas that free actions have proximate causes and that if an agent freely A-s at t in world W, he does not A at t in some other possible world with the same laws of nature and the same past up to t. Now, approximate causes of actions, including actions that are decisions, are internal to agents. In light of the general point about the proximate causation of actions, typical event-causal libertarianism encompasses a commitment to what may be termed agent-internal indeterminism. What I call modest libertarianism (see Mele 1995, pp. 211-21) embraces that commitment, too, even though it rejects the idea that the proximate causes of free actions indeterministically cause the actions. Indeterministic worlds in which every instance of causation within any agent is deterministic are hostile environments for libertarian freedom. What libertarians want that determinism precludes is not merely that agents have open to them more than one future that is compatible with the combination of the past and the laws of nature, but that, on some occasions, which possible future becomes actual is in some sense and to some degree up to the agents. The want something that seemingly requires that agents themselves be indeterministic in some suitable way - that some relevant things that happen under the skin are indeterministically caused by other such things. The focus is on psychological events, of course (as opposed, for example, to indeterministically caused muscle spasms), and, more specifically, on psychological events that have a significant bearing on action. Requiring internal indeterminism for free action and moral responsibility is risky. To be sure, quantum mechanics, according to leading interpretations, is indeterministic. But indeterminism at that level does not ensure that any human brains themselves sometimes operate indeterministically, much less that they sometimes operate indeterministically in ways appropriate for free action and moral responsibility. One possibility, as David Hodgson reports, is that "in systems as hot, wet, and massive as neurons of the brain, quantum mechanical indeterminacies quickly cancel out, so that for all practical purposes determinism rules in the brain" (2002, p. 86). Another is that any indeterminism in the human brain is simply irrelevant to free action and moral responsibility. Modest libertarians join other event-causal libertarians in taking this risk. In principle, an agent-internal indeterminism may provide for indeterministic agency while blocking or limiting our proximal control over what happens only at junctures at which we have no greater proximal control on the hypothesis that our universe is deterministic. Obviously, in those cases in which we act on the basis of careful, rational deliberation, what we do is influenced by at least some of the considerations that "come to mind" — that is, become salient in consciousness — during deliberation and by our assessments of considerations. Now, even if determinism is true, it is false that, with respect to every consideration — every belief, desire, hypothesis, and so on — that comes to mind during our deliberation, we are in control of its coming to mind, and some considerations that come to mind without our being in control of their so doing may influence the outcome of our deliberation. Furthermore, a kind of internal indeterminism is imaginable that limits our control only in a way that gives us no less proximal control than we would have on the assumption that determinism is true, while opening up alternative deliberative outcomes. (Although, in a deterministic world, it would never be a matter of genuine chance that a certain consideration came to mind during deliberation, it may still be a matter of luck relative to the agent's sphere of control.) As I put it in Mele 1995, "Where compatibilists have no good reason to insist on determinism in the deliberative process as a requirement for autonomy, where internal indeterminism is, for all we know, a reality, and where such indeterminism would not diminish the nonultimate control that real agents exert over their deliberation even on the assumption that real agents are internally deterministic — that is, at the intersection of these three locations — libertarians may plump for ultimacy-promoting indeterminism (p. 235). Modest libertarians try to stake out their view at this intersection." One kind of possible deliberator may be so constituted that no beliefs and desires of his that are directly relevant to the topic of his current deliberation have a chance of not coming to mind during his deliberation, whereas it is causally open whether some of his indirectly relevant beliefs and desires will come to mind. The causally open possibilities of this kind do not need to be extensive to secure the possibility of more than one deliberative outcome. Modest libertarians both need and fear internal indeterminism, and they are disposed to constrain it when engaged in the project of inventing indeterministic agents who can act freely and morally responsibly. The modest indeterminism at issue allows agents ample control over their deliberation. Suppose a belief, hypothesis, or desire that is indirectly relevant to a deliberator's present practical question comes to mind during deliberation but was not deterministically caused to do so. Presumably, a normal agent would be able to assess this consideration. And upon reflection might rationally reject the belief as unwarranted, rationally judge that the hypothesis does not merit investigation, or rationally decide that the desire should be given little or no weight in his deliberation. Alternatively reflection might rationally lead him to retain the belief, to pursue the hypothesis to give the desire significant weight. That a consideration is indeterministically caused to come to mind does not entail that the agent has no control over how he responds to it. Considerations that are indeterministically caused to come to mind (like considerations that are deterministically caused to come to mind) are nothing more than input to deliberation. Their coming to mind has at most an indirect effect on what the agent decides, an effect that is mediated by the agent's assessment of them. They do not settle matters. Moreover, not only do agents have the opportunity to assess these considerations, but they also have the opportunity to search for additional relevant considerations before they decide, thereby increasing the probability that other relevant considerations will be indeterministically caused to come to mind. They have, then, at least sometimes, the opportunity to counteract instances of bad luck — for example, an indeterministically caused coming to mind of a misleading consideration or, a chance failure to notice a relevant consideration. And given a suitable indeterminism regarding what comes to mind in an assessment process, there are causally open alternative possibilities for the conclusion or outcome of that process. Compatibilists who hold that we act freely even when we are not in control of what happens at certain specific junctures in the process leading to action are in no position to hold that an indeterministic agent's lacking control at the same junctures precludes free action. And, again, real human beings are not in control of the coming to mind of everything that comes to mind during typical processes of deliberation. If this lack of perfect proximal control does not preclude its being the case that free actions sometimes issue from typical deliberation on the assumption that we are deterministic agents, it also does not preclude this on the assumption that we are indeterministic agents. Now, even if garden-variety compatibilists can be led to see that the problem of luck is surmountable by a libertarian, how are theorists of other kinds likely to respond to the libertarian position that I have been sketching? There are, of course, philosophers who contend that moral responsibility and freedom are illusions and that we lack these properties whether our universe is deterministic or indeterministic — for example, Richard Double (1991) and Galen Strawson (1986). Modest libertarians can also anticipate trouble from traditional libertarians, who want more than the modest indeterminism that I have described can offer. Clarke, who has done as much as anyone to develop an agent-causal libertarian view, criticizes event-causal libertarianism on the grounds that it adds no "positive" power of control to compatibilist control but simply places compatibilist control in an indeterministic setting. Of course, given that combining compatibilist control with indeterminism in a certain psychological sphere was my explicit strategy in constructing a modest libertarian position (Mele 1995, pp. 212-13, 217), I do not see this as an objection. In any case, traditional libertarians need to show that what they want is coherent. That requires showing that what they want does not entail or presuppose a kind of luck that would itself undermine moral responsibility. The typical libertarian wants both indeterminism and significant control at the moment of decision. That is the desire that prompts a serious version of the worry about luck I sketched earlier. In the absence of a plausible resolution of the worry, it is epistemically open that a modest libertarian proposal of the sort I sketched is the best a libertarian can do. Of course, even if I happen to hit on the best libertarian option, it does not follow that I have hit on the best option for believers in free action and moral responsibility — as long as compatibilism is still in the running.
Mele's Models for Free WillBeing a self-proclaimed "agnostic" on these questions, Mele has developed both compatibilist and libertarian positions. His position on compatibilism needs some explaining. He says that because contemporary compatibilists (he mentions especially John Martin Fischer attend to what modern quantum physics tell us, the overwhelming majority do not believe that determinism is true. One might ask what they think free will is compatible with, if not determinism. The answer is that they believe that even if determinism were true, it would leave it open that people sometimes act freely. "Freely" here is in the compatibilist sense of free will that Immanuel Kant called a "wretched subterfuge and William James called a "quagmire of evasion." Mele says this is the traditional framing of the problem of whether "free action" (to be distinguished from free will") is precluded by determinism. Ever since Hume, as long as an agent is not coerced physically, her/his actions could be judged to be free, even if they are part of a deterministic causal chain. Mele's 1995 Modest Libertarianism discussed above is Mele's strongest two-stage model. In 1996 he developed a related position called "Soft Libertarianism," useful in the context of Frankfurt-style cases. Then in his 2006 work he developed a variation called "Daring Soft Libertarianism." Soft libertarians find determinism unacceptable because it claims that for all their intentions, evaluations, decisions, and subsequent actions, events were in progress before they were born that cause all those intentions and actions. This is the core concern of Peter van Inwagen's Consequence Argument. Soft libertarians are not primarily motivated because indeterminism may provide the alternative possibilities that are denied by Frankfurt cases, but simply that the causal chain of determinism might be broken, allowing them to make a causal contribution. He says,
Unlike hard libertarians, soft libertarians leave it open that determinism is compatible with our actions' being up to us in a way conducive to freedom and moral responsibility [presumably in the second stage of a two-stage model?]. However, they believe that a more desirable freedom and moral responsibility require that our actions not be parts of the unfolding of deterministic chains of events that were in progress even before we were born. If soft libertarians can view themselves as making some choices or decisions that are not deterministically caused or that are deterministically caused by, for example, something that includes deliberative judgments that are not themselves deterministically caused, then they can view themselves as initiating some causal processes that are not intermediate links in a long deterministic causal chain extending back near the big bang.Soft libertarianism differs from modest libertarianism in that it does not require robust alternative possibilities. But, somewhat inconsistently?, Mele says (p.113) that soft libertarians do not assert that free action and moral responsibility require the falsity of determinism. Mele briefly mentions a "soft compatibilism," but does not develop it beyond saying that "soft compatibilism leaves soft libertarianism open but is not committed to it." In his 2006 book Free Will and Luck, Mele extended his soft libertarian idea to "Daring Soft Libertarianism." Mele reaches out to Robert Kane's idea of Ultimate Responsibility, in which we can be responsible for current actions, ones that are essentially determined by our character and values, as long as we formed that character ourselves by earlier free actions that he calls Self-Forming Actions (SFA). SFAs in turn require brains that are not deterministically caused by anything outside the agent. Some may argue that a modest libertarianism gives libertarians all the openness they can get without introducing into an agent a kind of openness that entails freedom-precluding and responsibility-precluding luck. But libertarians like Kane will not settle for such modest libertarianism. For them, Mele developed a more daring soft libertarian view, DSL. Daring soft libertarians, he says, especially value a power to make decisions that are not deterministically caused - a certain initiatory power. They opt for event-causal soft libertarianism (p.113). They do not like decisions made indeterministically or at random, what Mele calls basically free action. But they accept what Mele calls basically* free action (Note the asterisk), whose requirement for alternative possibilities at the time of action are reduced, but whose requirement for indeterministic free actions some time in the past (Kane's SFAs?) is intact (p.115). They can then replace the indeterministic connection between judgments and actions with a deterministic one (p.117). (Note this can only be the adequate determinism of the two-stage models like Mele's modest libertarianism.) Mele says that,
Part of what DSLs are driving at in their claims about influence is that probabilities of actions — practical probabilities — for agents are not always imposed on agents. Through their past behavior, agents shape present practical probabilities, and in their present behavior they shape future practical probabilities. The relationship between agents and the probabilities of their actions is very different from the relationship between dice and the probabilities of outcomes of tosses. In the case of dice, of course, the probabilities of future tosses are independent of the outcomes of past tosses. However, the probabilities of agents' future actions are influenced by their present and past actions.(p.122) DSLs maintain that in the vast majority of cases of basically* free actions and actions for which agents are basically* morally responsible, agents have some responsibility for the relevant practical probabilities... These chances are not dictated by external forces, and they are influenced by basically* free and morally responsible actions the agents performed in the past.
The Libet ExperimentsMele is a powerful critic of interpretations of the Libet experiments which claim our decisions are made in our brain long before our conscious awareness of the decisions. He criticizes the interpretation of the Libet results on two grounds. First, the mere appearance of the RP a half-second or more before the action in no way makes the RP the cause of the action. It may simply mark the beginning of forming an intention to act. In the two-stage model, it is the considering of possible options. Libet himself argued that there is enough time after the W moment (a window of opportunity) to veto the action, but Mele's second criticism points out that such examples of "free won't" would not be captured in Libet experiments, because the recording device is triggered by the action (typically flicking the wrist) itself. Thus, although all Libet experiments ended with the wrist flicking, we are not justified in assuming that the rise of the RP (well before the moment of conscious will) is a cause of the wrist flicking. Libet knew that there were very likely other times when the RP rose, but which did not lead to a flick of the wrist, so his experiment could not detect them. Thinkers (e.g., Daniel Wegner, Patrick Haggard) who claim that the Libet experiments prove that our conscious will and subsequent actions are caused by prior neural activity - this is the popular view that "my neurons made me do it" - are simply wrong.