1. A PUZZLE ABOUT AESTHETIC RISK

It is common for artists to emphasise the importance of risk to their work. Think about an improvisation played by a jazz musician. It may be important to the quality of the work that the musician is taking a number of risks, such as in terms of the degree of creativity involved, or the kind of skill on display. The musician might well talk about playing ‘on the very edge’, and indeed offer this as an example of why her performance of the piece in question was successful. Moreover, the audience might well value the performance very differently because they recognise that the artist is taking risks in this way. ‘Playing safe’ does not seem to be a good recipe for great art, at least of this variety anyway, whether from the viewpoint of the artist or her audience.

There are many similar cases that spring to mind. Consider, for example, the performance of lead actor in a drama. The actor may well need to infuse the role with energy and emotion, while all the while aware of the danger that her performance might descend into melodrama. In pushing the dramatic limits of her performance, she is thus taking an aesthetic risk. Moreover, as with our jazz musician, this risk can be vitally important to the value of the performance, in that a parallel performance that lacks this dimension of risk may well be dull and pedestrian. Note too that while these are both cases of individuals taking aesthetic risk, we can also extend the line of reasoning in play to collaborative aesthetic enterprises. It may well be, for example, the whole theatre company that is taking an aesthetic risk in putting together a particular play and presenting the work in the specific manner that they do.Footnote ¹

The risk in question in these scenarios certainly seems genuine, but it poses a puzzle. When we talk of risks, we have in mind a particular scenario that would obtain if this risk were realized. Call this the risk event. In the case of the jazz musician, for example, this could be the risk event of technically messing up the piece, or the risk event of being too improvisational such that the piece loses its aesthetic appeal. Indeed, typically there will be multiple kinds of risk event in play, along several difference axes of evaluation (i.e. the artist is taking risks along several fronts simultaneously). In order to keep our discussion as straightforward as possible, however, we will henceforth focus on one particular kind of risk event associated with the target instance of aesthetic risk.Footnote ²

Here is the puzzle. It is very natural to think about risk in terms of the probabilistic likelihood of the relevant risk event. If it is likely (i.e. high probability) that this risk event should obtain, given the activity in question, then the activity is risky; whereas if it is unlikely (i.e. low probability) that this risk event should obtain, given the activity in question, then it is not risky.Footnote ³ This is the probabilistic account of risk, and it is the standard account of risk in the literature.Footnote ⁴ So, for example, juggling sticks of dynamite is a high-risk activity, since it is very likely that the target risk event (i.e. being blown to smithereens) will occur. In contrast, if one considers an activity like taking a quiet stroll in the park, the likelihood of this same target risk event obtaining is very low. It is thus, in this regard anyway (and in comparison with juggling sticks of dynamite), a low-risk activity.

The problem is that when we apply this very natural way of thinking about risk to certain cases of aesthetic risk, then it seems to generate entirely the wrong result. In order to see this, imagine that our jazz musician is really one of the greats, and for the sake of argument let us stipulate that the target risk event when she takes aesthetic risks concerns playing beyond her technical capacities (and so making mistakes that ruin the performance). It certainly seems right to say that in pushing herself to the very edge of her technical abilities in order to create this performance she is taking a genuine aesthetic risk. But given that she really is one of the very best jazz musicians – i.e. she is someone who has an exemplary level of technical expertise – do we really want to say that it is likely that the target risk event will obtain?

Indeed, notice that this musician will regularly describe herself as taking aesthetic risks of this kind – and her audiences will no doubt concur – but will rarely, if ever, actually make any serious technical errors as a result. In fact, we can even stipulate that this is the case – i.e. that despite her apparent risk-taking, she has never (as a mature performer, say), pushed herself too far and thereby messed up the performance. But if she regularly performs in this fashion, and the risk event doesn't obtain, then that surely entails that this risk event has a low probability of occurring. According to the probabilistic account of risk, therefore, she is not in fact taking an aesthetic risk here at all, despite her claims (and the claims of her audience) to the contrary. Insofar as the presence of aesthetic risk is contributing to the value of the performance, it thus follows that this performance is not as valuable as we hitherto supposed.

In summary, the puzzle arises out of a conflict between the following two claims:

1. (1) There are certain paradigm cases of aesthetic risk where the target risk event is unlikely (i.e. it is a low probability event).

2. (2) For an activity to be risky, the target risk event must be likely (i.e. a high probability event).

With the puzzle so posed, two natural responses suggest themselves. The first is to argue that we shouldn't take such cases of apparent aesthetic risk at face-value. That is, one might argue that if it really is true that, say, our great jazz musician is unlikely to technically mess-up her performance, then she isn't really taking an aesthetic risk at all. As a possible diagnosis of what is going on here, we might maintain that it only seems that this is aesthetically risky because we are imagining someone less technically proficient playing so close to the edge of their abilities. This imagined case would be a genuine case of aesthetic risk, since it would be likely that the risk event obtains. But once we remind ourselves of the great technical skill of the actual musician in front of us, we should realize that she isn't really taking any risks at all (at least along this technical axis of evaluation anyway).

This response to the puzzle effectively reaffirms (2) at the expense of (1). In contrast, I want to suggest a very different response to this puzzle, one that instead reaffirms (1) at the expense of (2). The problem, I claim, is not with these cases of aesthetic risk, but rather with the particular way of thinking about risk offered by the probabilistic account. That is, these cases of aesthetic risk are genuine; it is the account of risk that is on offer that is faulty.

2. CONTRA THE PROBABILISTIC ACCOUNT OF RISK

Rejecting the probabilistic account of risk might initially seem foolhardy. In particular, note what such a rejection must mean in this case (i.e. such that it can rescue the cases of aesthetic risk noted above). What we are after, it seems, is a way of thinking about risk such that an activity (e.g. a jazz performance) where the target risk event is unlikely (i.e. low probability) can nonetheless be genuinely risky. That may well appear to be a tall order. Nonetheless, this is precisely what I will be defending.

A consequence of the probabilistic account is that, keeping all other factors equal, two activities where the risk event associated with those activities are equally likely will be subject to the same degree of risk. Consider again our example pairing from earlier, of juggling dynamite and having a quiet stroll in the park, where the target risk event in both cases is identical (i.e. being blown to smithereens). If it turns out that the likelihood of the target risk event if one is having a quiet stroll in the park is in fact as high as it would be if one were juggling dynamite, then on this view it really would be just as risky. And that sounds right. For consider what it would be for it to be likely that taking a quiet stroll in the park would lead to one being blown to smithereens. One would need to imagine, for example, that the local park is in fact littered with land mines that could be set off at a moment's notice were you to step on them. If that's what a quiet stroll in the park is actually like, then it's no wonder that it's as risky an activity as juggling dynamite.

Here is the crux, however. For while the level of risk generally tracks the probability of the risk event in this way (at least provided we keep all relevant factors, such as the nature of the risk event, fixed), this is not universally the case. Indeed, we can easily imagine pairings of analogous scenarios with identical risk events, and with identical probabilistic likelihoods of the risk event obtaining, where the two scenarios are not subject to the same level of risk. Here is one such pairing:

Note that the probability of the risk event in each case is by stipulation identical (14 million-to-one) and so the probabilistic account of risk is committed to treating the risk in play in each case as being of an identical level. I contend, however, that there is a much higher level of risk in play in case 1 than in case 2.

In order to see this, we only need to note that in case 1, even despite the very long odds involved, the bomb blast is nonetheless something that could very easily occur. All it would take for the bomb to go off, after all, is that a few coloured balls in the lottery draw fall in a certain configuration. I don't think anyone who knew about this bomb plot would be sitting comfortably while watching the next lottery draw, since there is a serious risk that many people will soon die from a large bomb blast.

Would there be any corresponding cause for alarm in case 2, however? I claim not. Each of the three events, while the kind of thing that could potentially occur, is incredibly far-fetched. That is, none of them are events that could very easily occur. For all three to obtain would require an incredible run of events. That's not to say that there is no risk of the bomb going off, since all three of these events are genuine possibilities – as we might say, stranger things have happened. But the point is that the possibility that the bomb goes off in case 2 is not something that could very easily occur in the way that it is in case 1, even despite the sameness of the probabilities involved. As we might put the point, while both risk events are probabilistically unlikely, only the risk event in case 1 is modally close.

This might be initially puzzling, since if the probability of the risk event occurring is the same in both cases, then insofar as one of these risk events could very easily occur, shouldn't they both be possibilities that could very easily occur? Indeed, the proponent of the probabilistic account could be forgiven for thinking that the right response to this puzzle is to insist that the probabilistic likelihood of the risk event occurring, and its modal closeness, must go hand in hand. Accordingly, either we are wrong to suppose that one of the two risk events is modally closer than the other, or else we are wrong to suppose that they are equally likely. On the first option, one can preserve the thought that both scenarios are equally risky. On the second option, one can grant that the two scenarios are not equally risky, but this is no longer a counterexample to the probabilistic account. Given that determining the probability of an event obtaining in a real-world scenario is notoriously hard, one suspects that the proponent of the probabilistic account would be tempted by the second option. That is, couldn't they simply argue that the difference in modal closeness of the two risk events indicates that we must have a risk event in case 2 that has longer odds than its counterpart in case 1?

While estimates (or stipulations, in our case) of probability can always be disputed, unfortunately for the probabilistic account it is fairly straightforward to show that there must be pairings of cases where the probabilistic likelihood of the risk event comes apart from its modal closeness. The point is that we naturally order possible worlds, and thus the possible events that obtain in those worlds, in terms of their similarity to the actual world, where similarity is determined by how much needs to change in the actual world in order to get to this possible world where the target event occurs. A close possible world is thus one that does not require much change in the actual world. A far-off possible world, in contrast, is one that does require a great deal of change in the actual world. So there is a close possible world in which all that is different is that the cup on my desk is positioned slightly to its left. In contrast, the possible world in which Paris has switched places with Tokyo is much more distant, modally speaking. The first possibility is thus something that could very easily occur – it is an easy possibility – whereas the latter is not something that could easily occur, in that a great deal of change separates the actual world from this possible world.Footnote ⁵

Here is the nub of the matter. Although in general close possible worlds will tend to be worlds where high probability events occur, and far-off possible worlds will tend to be worlds where low probably events occur, there are exceptions. In particular, there can be close possible worlds where very low probability events occur – i.e. where such events are easy possibilities, even despite their low odds of obtaining. Indeed, lotteries are a classic way of illustrating this point. Although a lottery win is an event with very low odds of obtaining, it is nonetheless an easy possibility, since not a lot needs to change about the actual world for one to be a lottery winner – just a few coloured balls need to fall in a slightly different configuration.

Indeed, this is why people play lotteries. In particular, it is why people play lotteries even though they would not generally place bets on events obtaining that have astronomical odds. This is because events with astronomical odds are usually events that only obtain in far-off possible worlds, and one would be unwise to bet on an event of this kind obtaining. In contrast, a ‘bet’ on a lottery win, while also having astronomical odds, is nonetheless a bet on a possibility that could easily obtain – i.e. it is a bet on an easy possibility.Footnote ⁶

Going back to the puzzle we posed for the probabilistic account of risk earlier, we can see why it would be hopeless for the proponent of this view to try to evade the puzzle by arguing that the probability of the risk event in case 2 is in fact much higher than the probability of the risk event in case 1. For even if they manage to make this claim stick in this particular pairing of cases, it ought to be clear that the underlying problem is not thereby resolved. For there certainly must be pairings of cases with just this structure – i.e. where the target risk events are equally likely, but where one of the risk events is modally closer than the other. For the same reason, there is no mileage in pursuing the other option of insisting that if the risk events are equally likely, then they must be equally modally close.

This means that the only response left open to the proponent of the probabilistic account is to contend, against intuition, that the risk events at issue in cases 1 and 2 are equally risky, even though they are not equally modally close. Presumably, the line of argument will be that although there is a modal difference between these two events, in that the one is an easy possibility whereas the other isn't, this is not a difference that has any bearing on the objective risk in play. Instead, this is solely a matter of the probability of the event obtaining, and since this is the same in both cases, so both events are of equal risk.

While this approach is undoubtedly a theoretical option in this regard, it is clearly heavily theory-driven. Unless we had already signed up to the probabilistic account of risk, why would we be at all tempted by this line of argument? In particular, notice that the point that low probability events can nonetheless be modally close is not itself the product of theory, but rather rooted in our everyday assessments of events. We have just noted this point with regard to playing the lottery (i.e. in that people who play lotteries aren't similarly inclined to bet on modally far-off events with similar odds), but this isn't the only support available for this claim. Indeed, it is widely noted in the empirical literature on the psychology of risk ascriptions that our judgements about risk are primarily responsive to the perceived modal closeness of the event and not to its probability.Footnote ⁷ In particular, subjects will grant that two risk events might be equally likely to occur from a probabilistic point of view and yet nonetheless judge the one event to be riskier than the other. When this occurs, it is clear that the reason for this is that the subject regards the obtaining of the riskier event as modally closer than the other event.Footnote ⁸

The upshot of the foregoing is that we should replace a probabilistic account of risk with a modal account of risk, where the latter puts the focus on the modal closeness, rather than the probabilistic likelihood, of the risk event. Note that in adopting the modal account we are not thereby saying that the probability of the risk event is irrelevant to our judgements about risk. On the contrary, since modal closeness tends to go hand-in-hand with probabilistic likelihood (i.e. in the sense that the more likely the obtaining of that event is, the more modally close it is), making judgements about risk by considering the probabilistic likelihood of the risk event will often be a very sound way of proceeding. The crucial point, however, is that using this approach will also lead us astray in a particular range of cases where one has low probability risk events which are nonetheless easy possibilities.Footnote ⁹

3. THE PUZZLE RESOLVED

So, going back to the puzzle about aesthetic risk that I presented earlier, the claim that I am rejecting is (2) rather than (1). In particular, my contention is that once we have rejected (2) by replacing the probabilistic account of risk with the modal account, then we are in a position to reaffirm (1) without difficulty. How does this work? Well, consider again the cases of aesthetic risk that troubled us. Recall that they essentially involved low probability risk events, but which nonetheless seemed to be genuine cases of aesthetic risk. We are now in a position to see what is going on here. The great jazz artist who regularly plays right to the limit of her technical ability, but who nonetheless manages rarely, if ever, to exceed those limits (and thus fail) is nonetheless taking a large risk. After all, while it is not likely, given her track-record of success, that she will fail, it is at the same time true that she could very easily fail. That is, while the risk event of her failing in this performance is not probabilistically likely, it is nonetheless an easy possibility. This is just the point of saying that she is playing right to the very limit of her technical ability. So, expressed, the idea is that very little would need to be different for her accidentally to cross that limit, leading to the obtaining of the risk event.

And what goes for great jazz artists and the limits of their technical expertise applies to analogous cases of aesthetic risks, and different axes of evaluation along which we might determine those risks. The brilliant actor who, night after night, manages to stay just the right side of melodrama in her performance is taking an aesthetic risk. For although the target risk event of her slipping into melodrama is unlikely, given her success rate, it is something that could so easily occur. The general point is that once we have the right account of risk in mind, then, we can comfortably make sense of the idea of there being aesthetic risks of just this kind.Footnote ¹⁰