The tough choices argument
Recently, E. J. Coffman (Reference Coffman2011) has argued that an omniprescient being – one who always believes, with maximal confidence, all and only truths, including truths about the future – is capable of making ‘tough choices’, and, hence, can intend and act intentionally. He reasons by way of rebutting what he calls ‘the tough choices argument’, one claimed to be derived from argumentation in Kapitan (Reference Kapitan1986, Reference Kapitan1991).
On the definition of a tough choice as a choice between alternative courses of action none of which is known by the agent to be uniquely reasonable, here is the tough choices argument (TCA). Suppose, for conditional proof, that:
-
1. S is omniprescient.
Add this critical premise:
-
2. If S makes a tough choice between alternative courses of action A and B, then (prior to the choice) S believes about each of A and B that her taking it is epistemically possible [that is, S believes about each of A and B that she may take it (where ‘may’ here expresses an epistemic, as opposed to a metaphysical, modality)].
It follows that,
-
3. If S believes about each of A and B that she may take it, then S lacks a maximally firm occurrent belief as to whether she will take A.
So,
-
4. If S makes a tough choice between A and B, then (prior to the choice) S lacks a maximally firm occurrent belief as to whether she will take A.
But,
-
5. S has a maximally firm occurrent belief as to whether she will take A.Footnote 1
So,
-
6. S does not make a tough choice between A and B.
So,
-
Conclusion: If S is omniprescient, then S does not make a tough choice between A and B.
The conclusion is significant for those with some stake in debates about divine agency, and Coffman notes that ‘most traditional theists should think that God makes some tough choices’, notably, in creating the world.
Coffman offers two main considerations to undermine the motivation for premise (2) of the TCA. First, he offers counterexamples to a general principle from which (2) can be derived, and second, he contends that a rival principle works equally well in explaining the relevant data. Let us consider these in turn.
Foreknowledge of one's intentions
Coffman's first consideration begins by noting that premise (2) can be supported by the following claim:
(KP): If a person acquires an intention to (take course of action) A, then (prior to gaining the intention) the person believes both that she may A and that she may not A – i.e., prior to gaining the intention, she believes about each of A and not-A that it is epistemically possible.
However, as Coffman points out, while decisions are partly prompted by uncertainty about what to do, not all intentions are formed by way of decision and not all intention-acquisition presupposes prior uncertainty. Suppose I arrive at my office door and acquire an intention to unlock it with my key without prior deliberation. I already knew that I would intentionally unlock the office door because this is my routine, and so I was neither previously uncertain nor viewed my performance as epistemically contingent. A similar thing occurs when I intentionally swerve my car to avoid hitting an animal that darts in front (Mele (Reference Mele2003), 200–201). Again, suppose you receive a call from a distraught friend who desperately needs to talk to someone about the sudden and unexpected death of a family member. In the past, you have always decided in favour of talking to your friend, for this is the kind of person you are, and consequently you know you will decide the same way today. While you haven't gained an intention to do so yet, you can gain that intention and eventually you do.Footnote 2
Coffman correctly observes that these cases do not directly threaten premise (2) of TCA since none involves a tough choice. But they are evidence that one can anticipate what one will do intentionally before acquiring the intention, thus, that its complement is not epistemically possible, contrary to what KP implies. He concludes that one source of support for premise (2) is weakened.
Coffman is correct to fault KP, but the situation is more complex than he acknowledges. There is a distinction between general intentions or policies to engage in a type of behaviour, say, unlocking one's office door with a key or talking to distraught friends in need (Bratman (Reference Bratman1987), 56–57, 87–91), and specific intentions, e.g., to unlock the door at a particular time. The latter divide into distal (future-direct) and proximal (present-directed) intentions. One may certainly anticipate a particular intentional act before it occurs because one sees it as an instance of one's policy or as a particular node in an action plan one has endorsed. So, while I need not view my unlocking the door now as epistemically contingent before acquiring the correlated proximal intention, my anticipation may derive from a previous general or distal intention which itself was antedated by correlated beliefs in epistemic contingency. If so, it may still be that when faced with a tough choice one must believe either that a particular alternative is epistemically contingent or that alternatives of that type are epistemically contingent. This modification of premise (2) is equally able to generate the TCA.
A rival account of the tough choice data
Turning to his second consideration, Coffman agrees that one faced with a tough choice between alternatives believes that taking either is possible, but rather than a belief in epistemic possibility of the sort required by premise 2, he thinks the relevant data can be handled by a different principle, namely, what he calls a Belief in Freedom Requirement:
(BFR) If a person makes a tough choice between alternative courses of action A and B, then the person believes about each of A and B that she's (metaphysically) free to take it (alternatively, that she has it within her power to pursue it).
Coffman does not specify what metaphysical freedom is. One might think that he takes a metaphysically free act to be not causally determined by antecedent circumstances, but his treatment of D3 (see below) allows for a compatibilist-friendly reading. This lack of specificity makes it difficult to determine what exactly BFR attributes.
Here are the data that Coffman finds to constitute ‘the best available prima facie case’ for premise (2) of the TCA:
(D1) If you make a tough choice between alternative courses of action A and B, then (prior to the choice) you believe that it is (in some sense) possible for you to take either of A and B.
(D2) If a person who's trying to make a tough choice is asked whether he's aware of factors that threaten the freedom of the upcoming choice, the following would be a natural reply: ‘No, I'm not aware of any such factors. So far as I can tell, the outcome is up to me.’
(D3) A Determinist (i.e. someone who believes in Determinism – the remote past and physical laws together fix a unique future) could make a tough choice without thereby holding explicitly inconsistent beliefs.
Now Coffman finds that BFR easily explains D1, and since it is not obvious that (metaphysical) freedom rules out determinism, then ‘somewhat surprisingly, BFR can honor D3’ (Coffman (Reference Coffman2011), 51–52). As for D2, Coffman notes that the agent's qualification of his own presumed freedom by ‘so far as I know’ is occasioned by the very question, ‘Are you aware of any freedom-threatening factors?’ Perhaps being asked the question causes the agent to suspend his belief in metaphysical freedom and retreat to the more qualified self-ascription, or, alternatively, to retain the belief but offer the qualification because there might unconsidered evidence to the contrary. If so, then the proponent of BFR can deliver a plausible account of D2 without recourse to presumptions of epistemic contingency (ibid., 52).
This reasoning is inconclusive. First, BFR is not obviously a rival to premise (2), for it may well be that one who takes himself to be free with respect to actions A and B thereby thinks that both are epistemically possible, or so I have argued (Kapitan (Reference Kapitan1986), 239–241; idem (Reference Kapitan1989), 31–34; idem (1991), 109–113). If premise (2) is entailed by BFR then the burden of explaining D2 is shouldered by (2) alone. Second, the claim that BFR can account for D3 succeeds only on a compatibilist account of metaphysical freedom, but unless a compatibilist account of presumed freedom is proposed and defended – which Coffman does not do – then his handling of D3 is a mere promissory note. By contrast, the argument in Kapitan (Reference Kapitan1991) is built on a compatibilist account of presumed freedom that not only accommodates D3, but entails premise (2).Footnote 3
The acquisitions of intentions
Here is an informal version of the argument that omniprescience precludes intentional agency as presented in Kapitan (Reference Kapitan1991), (Reference Kapitan1994), and (Reference Kapitan1997). Because
-
1. intentional action is caused by intentions;
-
2. an agent who acts intentionally acquires some intentions;
-
3. in the causal aetiology of any intention is an antecedent sense of options;
-
4. a sense of options includes a sense of epistemic contingency; and
-
5. a sense of epistemic contingency implies a sense of uncertainty:
then
-
6. an omniprescient agent cannot act intentionally.
More dramatically, an omniscient being is omni-impotent. Assuming (1) – a standard tenet of the causal theory of action – let me motivate (2)–(5) in turn.
The case for (2)
Intention is a practical psychological state of being settled upon either a particular course of action or type of action, distinguishable from innate behavioural tendencies by being an acquired state. Although the spontaneous motility of the foetus ‘provides the raw material from which goal-directed actions are selected’ (Adolph & Robinson (Reference Adolph, Robinson, Liben and Muller2015), 7), intentional behaviour is learned, not simply the manifestation of an innate propensity (von Hofsten (Reference Hofsten2007), idem (Reference Hofsten and Prinz2012), Adolph & Joh (Reference Adolph, Joh, Woodward and Needham2009) ). Intentions are initially acquired as a reaction to our experience of prenatal movements and their outcomes, and are subsequently refined and augmented in response to challenges posed by the juxtaposition of aversions, desires, and goals with presented opportunities or affordances (Adolph & Robinson (Reference Adolph, Robinson, Liben and Muller2015), 21–23). These responses are initially particularized, say, a specific grasping or sucking intention, but with time and experience patterns of behaviour, e.g. of looking, grasping, pointing, etc., are acquired.
Some intentions must be acquired if an agent acts intentionally. The reason is that to act intentionally is to relate oneself purposely to various particulars, whether one's own body parts or distinct persons, objects, events, places, and times, and intentions are action-guiding only if their contents direct our efforts within a realm of particulars. As such, an agent needs information about their properties, relations, and spatio-temporal positions (von Hofsten (Reference Hofsten2007), 55). If I intend to submit a paper to Religious Studies next week, what I intend includes reference to a particular agent (myself), a particular paper, journal, interval, and so forth. Only thus do I initiate a chain of events which constitutes my intentionally sending that paper to Religious Studies. Even if a general intention were innate, say, to help the needy, it would require a further specific intention to help that man over there and to do so now. There can be no innate propensity to expend intentional effort in just that direction at just this time, etc. The information is gained only through interaction with particulars, though not necessarily with those with whom one plans to interact. Even a creator could not anticipate its own creative actions without information about particulars, if only to its own imaginations, plans, expectations or intentions, themselves particular states. Specific intentions must be acquired because they are dependent upon acquired information.
The case for (3)
We deliberate about a course of action only if we take it to be an option. Integral to this attitude are feelings of efficacy (the action is something that we are likely to do should we try) and of contingency (viz., it is, as yet, both possible that we perform the action and possible that we do not). Both, in turn, are components of a typically tacit sense that the future is open with respect to such performance, a sense fully conceptualized in the reflective awareness we expect of an omniscient agent.Footnote 4
Does a sense of options accompany all instances of intention formation? To act intentionally is to expend effort, and to acquire an intention is to ready oneself for the expenditure of effort. It would be pointless to expend effort without sensing both a need for so doing in order to reach desired ends and at least a chance of its success (Preston & Wegner (Reference Preston, Wegner, Morsella, Bargh and Gollwitzer2009), 575–577). One source of evidence for this claim are psychological studies of ego-depletion, according to which there is a limited amount of energy available to an agent for voluntary action, energy that is depleted in the processes of decision and self-regulation (Vohs (Reference Vohs, Baumeister, Mele and Vohs2010), 70–71). Expectedly, agents tend to conserve limited energy, a propensity that some psychologists recognize through a principle of least effort, namely, that organisms tend to expend the least amount of energy in the achievement of a goal.Footnote 5
This principle should not be confused with a principle of universal laziness, or with a claim that agents always select the most efficient means; we waste energy because we misperceive what is needed. But the principle does support the idea that at least rational agents do not act voluntary except by sensing that there is both a need for a particular effort if a desired result is to be achieved as well as a chance that the effort is likely to pay off. By contrast, if one felt that the result would happen no matter what, or, alternatively, that it would not occur no matter what, then any effort to bring it about would be viewed as either pointless or futile. This is just another way of saying that an agent forms an intention to do an action A only on the assumption that both A-ing and not A-ing are options.
The case for (4)
What sort of contingency is involved in this presumption of openness? Most agree that mere logical or nomological possibility is not enough. An act is taken as possible relative to some set of particular circumstances or considerations, e.g. given the standing conditions, I can leave this room within the next minute but not visit the Tower of London within that span. Plainly, the cans, cannots, mights, and maybes of everyday usage, in whose terms we plan, predict the future, and ruminate about the past, express a modality indexed by a totality of particular circumstances, what we might express with ‘prevailing circumstances’ or ‘things as they stand’.
Some are tempted to hold that an action is taken as optional only if viewed as consistent with all past facts. But besides saddling agents with inconsistency should they have deterministic leanings, it is dubious that ordinary agents have sweeping beliefs about the entire past. A more modest proposal is to relativize the sensed possibility to a proper subset of the past. However, this won't do given that false beliefs about prevailing conditions often both inflate and restrict perceived options. A different approach describes the sense of possibility in terms of a cognitive modality – e.g. consistency with what one knows, believes, or accepts – noting that our cognitive limitations open up a wide sea of such epistemic or doxastic possibilities with ample room for deliberation. Rather than positing a sense of possibility relative to what is the case, this view relativizes possibility to what one takes to be the case or, more generally, to one's commitments, so that each feeling of ‘I can’ is qualified by an implicit ‘as far as I can tell’.
Suppose I firmly accept that tomorrow I will intend to return to London in the afternoon. Then, as far as I can tell, the future is settled concerning that action. If this expectation derives from my own practical commitments then I already have the intention to do so, or to act on an action-plan one node of which is returning to London in the afternoon, and there is no need to form the intention again. If I accept it because I think that this is the way the future will turn out regardless of what I do, then I do not sense a need to form that intention now as I do not feel that future is open as regards that action. I may realize that there are plenty of other things to do, say, to get up and get dressed, get myself to the train station, step onto that train then, etc., but the intentions to do so are fine-grained enough to be distinct from the intention of returning to London. The upshot is that intention, as well as knowledge and belief about what will happen, settles the future so that it is no longer perceived as open.
The case for (5)
It is but a small step to recognize that for suitably reflective beings, a sense of epistemic contingency also involves a sense of uncertainty about what will happen – that is, a sense that one lacks a firm belief or commitment concerning a potential future action. What is important is that these feelings of indecision and of uncertainty are attitudes with their own content, not merely conditions of indecision and uncertainty where negation dominates the attitudinal verbs. Thus, an agent need not be uncertain or ignorant about whether he or she will A in order to take A-ing as optional; he or she can remain committed to A-ing while deliberating about taking an incompatible action instead if, in so doing, previous commitments are overlooked. I might know that I will play tennis with my daughter at noon on Friday and not play golf with a friend, but then forget about the former and deliberate about the latter. In so far as I take golfing as optional I feel I can do so relative to what I accept, and thus, feel uncertain about what I will do. Yet, at the same time, I know very well what I will do. This situation can occur because one typically does not access all one's commitments; forgetting does not imply abandoning a commitment but merely removing it from activated working memory.
The derivation of (6)
Given bivalence, an omniprescient being cannot be uncertain about any proposition and, as such, can never have a sense of its own uncertainty. For suppose that such a being, S, has a sense of uncertainty about whether P. Then the propositional content of that sense would be either true or false. If true, then S would be uncertain about P, in which case S would not be omniprescient. If false, then, again, S would not be omniprescient since, by definition, an omniprescient being believes only truths. So, an omniprescient being cannot have a sense of uncertainty. Accordingly, by (5) it cannot have a sense of epistemic contingency, by (4) it cannot have a sense of an open future, by (3) it cannot acquire an intention, and by (2) it cannot act intentionally.
If this reasoning is sound, it follows that an omniprescient being cannot make tough choices. Less than omniscient theists can, however, and so they should.
