Key to the traditional conception of free will has been the idea of alternate possibilities. The idea of alternate possibilities is a rather vague one, and various definitions have been proposed in order to clarify what exactly is meant by the assertion that “Free will requires alternate possibilities.” Some definitions focus on the idea of constraints, saying that one has genuine alternate possibilities if one is not prevented from certain courses of action either physically or mentally. This idea may even be compatible with determinism. The kind of conception of alternate possibilities that I wish to address is a much more robust form. In this view, one has alternate possibilities if in a given situation and holding the past constant, a person or agent may act in two different ways. My view is that the problem with holding on to the idea of alternate possibilities as a requirement for free will is that it is incompatible, not necessarily with the facts of the world, but with the very idea of freedom (I will use the word freedom interchangeably with free will in this essay). The definition of freedom or free will as the ability to do otherwise is flawed.

In this paper I will examine how taking alternate possibilities to be an essential component of free will leads, in an indeterministic world, to arbitrariness in our decision making. I will begin by showing the arbitrariness inherent in an agent-causal account of free will, and then show how the problem developed in that view is also a criticism one can make of the non-determined view of AP free will.

While this essay will deal almost exclusively with a libertarian viewpoint (the idea that the world is not determined and that we have free will), it will be useful to define a determined world in order to refine our definition of a world which is not determined. A concise and accurate description comes from God, Foreknowledge, and Freedom ed. John Martin Fischer (Stanford University Press, 1989). A causally determined world is such that:

For any given time, a complete statement of the "real" facts of the world at that time, together with a complete statement of the laws of nature, entails every truth as to what happens after that time.

There are two common stances that libertarian theorists have taken in order to outline a free agent with alternate possibilities. The first view is that actions may have alternatives (and therefore be free) if they somehow originate in the agent, as opposed to following the normal event-causal chain of the world. This requires some explanation. Most views of the world ascribe to the belief that only past events cause current and future events, just like the determinist worldview outlined above. Libertarians who believe in agent-causal actions believe that this view only describes one kind of event or action. They believe that while the normal event chain is sufficient to describe a world without agents, when we start talking about people we have to introduce a second kind of causation: agent-causation. This kind of causation is meant to describe what happens when an agent performs a free action. It is important to keep in mind that this kind of causation is not just like regular event causation. It is not that the past states of mind and body of the agent, along with drives, beliefs, etc., cause the agent to act the way she does in the same way that the velocity and direction of a club along with wind conditions determine where a golf ball will land. Rather, when we talk about an agent-caused action, we are talking about an event that begins with a particular kind of substance: the agent. Most causal views see a chain of events, the one causing another; in agent causation, a chain of caused events can begin with an agent. Another important facet of agent-causalism to remember is that agent causalists are not committed to saying that every action an agent performs is actually agent-caused. It is very possible for an agent to perform deterministically caused actions. It is free actions that must always be agent-caused in this view. There are immediate metaphysical concerns that spring to mind when one lays out this view, such as the nature of the agent, but they are not under scrutiny in this paper. Our goal is to analyze whether or not the agent causal view, if correct, gives us an accurate depiction of free will. I contend that it does not.

We now begin to see in what way a traditional agent-causalist would regard an action as free. An event is freely caused if and only if an agent directly caused the event itself or the causal chain that terminates in the event. This seems complicated, but is actually very similar to what I believe is a natural, non-reflective conception of free will. In our golf ball example, the velocity of the club, etc. determined the final location of the ball, but the decision to hit the ball is a free choice, with its origins in the agent. If the agent causalist is correct, then the free choice to hit the golf ball cannot have been entirely predictable by past states of the world and principles of nature. It must have been something in the agent, in that substance, that caused the choice.

There are many opinions on what exactly is the free action that an agent takes. Some say a choice (I will hit the golf ball now), some say an inclination prior to choosing (I feel like hitting that golf ball), and some say the entire event, from presentation of the choice to the agent to the beginning of the act decided upon (There is a golf ball. Should I hit it? Yes. Thwack!) What all traditional agent causalists agree upon is that whatever event is free must be caused by an agent, or in other terminology, an unmoved mover. This last term is meant to make explicit the point that the agent is not bound by causal laws to perform certain actions. She may be swayed by reasons and events towards one action or another, but nothing can make her do one over the other. This way she causes events (moves) without being tied to causal laws (unmoved).

What it means for an agent to cause something without being caused to do so is still a bit mysterious, so we will look at a specific example. Let us take a specific agent Lynn living in a world with agent causation. She has to make a choice between two shirts to wear for the day; one blue, and one red. If we really want to understand her choice our first move is to look at reasons. Perhaps she really likes red, and that is a reason one way. On the other hand, perhaps she wore red yesterday, and that fact inclines her towards the blue. A causal determinist would say that these reasons, along with all of the other facts about Lynn and the world, cause her to make one choice or another. A traditional agent causalist rejects this picture altogether, if Lynn's choice is to be a free one. He would say that unlike the flying golf ball and wind, reasons for one or another choice do not have causal power over an agent. They may influence the likelihood of a certain outcome, but they can never cause one or another. Now, I think, we are beginning to see why agent causation is very appealing. In any given situation, not only does one have the possibility to do otherwise (because reasons don't determine the decision), but also the action originates directly in the actor, despite being tempered by reflection upon experience and the nature of the agent's character.

Unfortunately for this view of causation, it contains a fatal flaw. Let us sketch it a bit by imagining two different worlds. The two worlds are identical, and they both contain Lynn, making her decision about which shirt to wear in the morning. In one world Lynn chooses to wear her red shirt, and in the other she wears the blue. The past world, before the choice was made, is identical in each world. Not only that, but Lynn herself is the same in each world, too. The problem is that there is nothing in the nature of the world or in Lynn's nature to explain the difference in choice. Each world was identical, by stipulation, until Lynn made her choice. The power of the agent causalist explanation is supposed to be that it explains free choices in terms of the agent, but here we see that even that feature seems to be suspect. Nothing about the agent, no reasons or inclinations or anything else, can explain the discrepancy between two 'free' choices. Her decision now seems arbitrary. Agent causal AP free will therefore fails this important test of freedom.

The second popular libertarian view is one that salvages several appealing aspects of the agent-causal view while getting rid of the troublesome second type of causation. It is not Lynn's nature as an agent that allows her to be free. Rather, it is that her choice is not determined, together with certain facts about that choice. The key term in this kind of view is “probabilistic causation”. Recall that a causal determinist believes that given a certain state of the world and fixed principles of nature, any future state of the world is fixed. Any one state of the world guarantees that a certain future state of affairs will come about. If a causal determinist believes that past states predict that a certain future state of affairs will always occur, a probabilistic causalist would say that that past states of affairs make it likely that a certain future state will occur. Our golf ball example might help here. Both the determinist and the agent-causalist would say that the velocity and makeup of the golf club and ball, along with the wind conditions etc., determine precisely where the ball will land. Someone who believes in probabilistic causation would say that even taking every factor into account, the final location of the golf ball can only be stated relatively. It might go 100 meters, or it might go 99. It might even go 1 meter, although the possibility of that is very low. We can at most say how likely it is that the golf ball will land in a particular area.

One popular instance of this point of view can be found in Robert Kane's description of free action. Kane takes the determinist view of action, and changes causal determinism to probabilistic causation. In a determinist view, Lynn's choice of shirt was causally determined by past events, mental states, etc. There was a set path along which she was led, and her choice was determined by the conditions surrounding the situation, including her mental states and other states of the world. Kane takes essentially this structure and replaces deterministic causation with probabilistic causation. While all of the same past mental states lead up to the decision, and the decision is based on these past states, Kane feels that for the decision to be free, it must be non-determined. What this means is that before the decision was made, given exactly the same conditions, it must have been possible that the agent choose something else. In this way, Kane's version of a free agent must exist in a libertarian (non-determined) world, and must have alternate choices available to her.

So far there is little to distinguish a choosing agent in Kane's view from our non-deterministic golf club swing. If we view the array of possible choices open to an agent like we view where the golf ball lands, it seems that little has changed by making causation probabilistic. The fact that an agent can choose one way or another does little to advance the case of free will. The choice is still caused by states of the world, and the choice that the agent eventually takes seems to be more or less arbitrary. An opponent of Kane's might say that all Kane has done is turned the agent into a pawn of a different kind of causation than deterministic. Even though different outcomes are possible in a situation, the outcomes seem to be caused not by the agent, but by the probabilistic laws governing the world. While Lynn may choose either the red or the blue shirt, it is still the case that it is her states of mind and the state of the world that cause her to do so; it is just that they do so in a probabilistic fashion, rather than a deterministic fashion.

Kane attempts to resolve this by appealing to what he calls a self forming willing (or self forming action; what he describes is both an action and a willing) or SFW. Kane describes these SFWs in the following way:

Actions are “will-setting” when the wills of agents (their motives and purposes) are not already “set one way” before they act…, but rather the agents set their wills one way or the other in the performance of the actions themselves. Choices or decisions are will-setting when they do not result from the agents' merely discovering during deliberation what they (already) favored, but when the agents make the reasons for preferring one option prevail at the moment of choice by choosing or deciding. Will-setting actions are in this sense “will-setting,” not already “will-settled.” (412)

One difference between Kane and agent-causalists is that he only views a handful of actions as directly free. He thinks that directly free actions are only those decisions preceded by a SFW, and he also thinks that SFWs occur only in cases where the agent is torn between decisions. According to Kane, it is these actions that create the nature of the agent; this is what he means by “self-forming” or “will-setting.” Therefore, there are many actions that are derivatively free, if they stem in part from a decision that was a result of a SFW. Since our situation of Lynn deciding which shirt to wear fairly obviously would not result in Lynn being torn between two choices, let us say that the blue shirt is a uniform for one job, and the red shirt the uniform for another. By wearing one or the other, she is deciding which job to accept; the decision is, therefore, full of conflict and self-forming.

Again, some clarification is needed to understand what Kane means by this self-forming willing. It might be tempting to classify it as a characteristic of an agent, but this is precisely what Kane wishes to avoid. A SFW is not enduring, but rather a product of an especially conflictive situation for the agent; indeed, it is an action that an agent takes. It is clear from the quote above that in a truly (not derivatively) free choice, Kane thinks that an agent does not just deliberate, but also performs this act, this SFW, the outcome of which then becomes a part of her personality. To help us understand what exactly he means by a SFW Kane gives us the analogy of a microphysical particle directed at a barrier. Because its velocity and mass cannot both be precisely known, we cannot know if it will penetrate a certain barrier. In the same way, we cannot know whether or not a SFW will succeed or not. Once the SFW fails or works, however, that decision becomes a special part of the agent, a part that is uniquely determined by the agent. Since the SFW is an action take by an agent, and it is the SFW that determines which choice the agent eventually attempts to carry out, the choice is therefore a part of the agent that, in a certain sense, the agent herself caused. For the agent causalist agents are unique substances, and express that uniqueness in free choice; for Kane, each SFW (and hence free choice) helps to determine what is most essential and unique to the agent.

A very important part of Kane's SFW is that it results in an action that is essentially agent influenced, and hence can be attributed to the agent (that is to say that the agent can be held responsible for the action), without agent-causalist “substance.” His criterion for this has three parts. First, the action made was made in accordance with the reasons the agent had for acting. Second, the agent could have acted in another way, also in accordance with the reasons the agent had for acting. Thirdly, the SFW was an action taken by the agent. Kane thinks that these three conditions for freedom create a complete picture of a free choice, one for which an agent is ultimately responsible. A problem that he fails to avoid is one of arbitrariness. The puzzle can be drawn as we drew the problem with agent causal alternate possibilities. Imagine two parallel worlds. They are identical, and our agent Lynn is choosing between shirts (and therefore jobs) in each of them. In both worlds, Lynn makes SFAs to decide to wear the red shirt. In one world her SFW succeeds, and in the other it fails. According to Kane, Lynn has made her choice, in each world, according to her own reasons. She has also made a choice in each world that will effect her development and lead her to develop in one way in the first world and another way in the second. Yet these differences arose from precisely the same set of circumstances. Not only was there nothing different about the world prior to her choice, but there was also nothing different about her, in any way. What, then, accounts for her developing as a different person in each world? Nothing, it would seem. Her difference in character, no matter how disguised in SFWs and reasons for action, is still arbitrarily decided in Kane's view. The fact is that the state of the world was identical until the point at which the choice was made, which means that nothing can account for the difference in choice from one world to the next.

Kane might respond that while each world is possible, one is more likely than the other. The facts of the world influence the SFW so that the first world might be 90% likely, or more or less than that. Thus the ultimate outcome is influenced by the facts of the world, and is not arbitrary. The problem with this response is that likelihood of one outcome or another does not address the initial problem. It is possible that either world occurs. Therefore, given both worlds, what accounts for the difference in the outcome of the SFW? Despite the fact that one choice may have been one percent likely and the other ninety nine, in order for their to be alternate possibilities as Kane sees them it must be possible for the one percent likelihood choice to have happened. Given that that world is so unlikely, we naturally look for some explanation as to why the least likely world occurred, but none is forthcoming. We are back to our to our non-determined golf ball. Just like we can predict the possibility of one landing spot for the ball over the other, we can predict the possibility of Lynn's choosing the red shirt over the blue. We cannot, however, account for why either the ball lands in one spot or another or why Lynn chooses one over another. That is the essential character of a non-determined world.

There is another route open to Kane. In the quote stated above, he says, “…the agents make the reasons for preferring one option prevail at the moment of choice by choosing or deciding.” Look at the agent in each world after the choice is made, he would say. In one world, a certain set of reasons prevailed, and Lynn chose the red shirt. In the other world, another set of reasons prevailed, and Lynn chose the blue shirt. In each world, there is a reason that one set won out over another. It is the SFW that she made in each choice that decided. In the one she has affirmed a certain set of reasons, and shaped her character in that way; in the other world, she has affirmed another set of reasons, and shaped her character in another way. Kane would say that nothing could be less arbitrary. The reasons in each case for choosing are her own. Not only that, it is an action that she herself undertook which caused one set or the other to prevail. It seems that the agent is fully responsible for the choice that she has made. It is simply evident that the agent has used her own reasons to make her own free choice, in each world. The response to this is identical in form to the earlier response. The fact remains that each world is identical up to the very point of choice, whereupon they split. If everything remains constant until that point, nothing after the fact can explain the difference in choice. The same reasons are present in the mind of the agent at the moment of choice, and the same deliberation has taken place. A SFW seems to be an arbitrary way of deciding which reasons ultimately take precedence, similar to rolling dice. The odds may be calculated, but in the end it all comes down to luck, and nothing more.

Alternate possibilities are very tempting for free will theorists, but they are irrational or arbitrary or both in practice. Both agent-causalist and libertarians in Kane's vein of thought fail to address how it is that one can have alternate possibilities that are not arbitrary.

O'Conner, Timothy. “Libertarian Views: Dualist and Agent-Causal Theories.” The Oxford Handbook of Free Will. Ed. Robert Kane. New York: New York, 2002. 338-354.

Kane, Robert. “Some Neglected Pathways in the Free Will Labyrinth.” The Oxford Handbook of Free Will . Ed. Robert Kane. New York: New York, 2002. 407-437.