A Companion to the Philosophy of Action

Timothy O'Connor, Constantine Sandis, Timothy O'Connor, Constantine Sandis

• Publisher: Wiley-Blackwell
• Year: 2012
• ISBN:
  9781118394243
• Edition: 1
• Topic: Philosophy
• Subtopic: Mind & Body in Philosophy

A Companion to the Philosophy of Action offers a comprehensive overview of the issues and problems central to the philosophy of action.

• The first volume to survey the entire field of philosophy of action (the central issues and processes relating to human actions)
• Brings together specially commissioned chapters from international experts
• Discusses a range of ideas and doctrines, including rationality, free will and determinism, virtuous action, criminal responsibility, Attribution Theory, and rational agency in evolutionary perspective
• Individual chapters also cover prominent historic figures from Plato to Ricoeur
• Can be approached as a complete narrative, but also serves as a work of reference
• Offers rich insights into an area of philosophical thought that has attracted thinkers since the time of the ancient Greeks

Part I
Acts and Actions

1

Action Theory and Ontology

E. J. LOWE

Any comprehensive theory of action should have something to say about the ontology of actions. It should address such questions as the following. What are actions, if indeed they are anything at all? – for we shouldn’t just assume that actions exist. Are they, for instance, a species of events? If so, then what are events, and what makes actions special among events? How are actions individuated and – if this is a different question – what are their identity conditions? Must every action have an agent (or agents) and, if so, what sort of thing can be an agent, and in virtue of what features can it be said to perform, or engage in, actions? In this chapter I shall say something about all of these questions.

What are Actions?

One obvious way to address this question is to look at action sentences and examine their apparent ontological implications. A typical action sentence would be ‘John opened the door.’ Here John is represented as having performed a certain type of action – opening a door – and thus is represented as having been the agent of a token action of that type. (I take it that the type/token distinction is too familiar to need further elaboration here.) By implication, this token action occurred at some specific time in the past. Extrapolating from this kind of example, we may venture to say that token actions are particular occurrences of certain action types, each possessing an agent (or agents) and a particular time of occurrence. In answer to the question ‘But do we really need to include token actions in our ontology?’ the following line of argument, due originally to Donald Davidson (1967), may be advanced. Action sentences such as ‘John opened the door’ can be adverbially modified in indefinitely many ways. For instance we can expand this sentence into one such as: ‘John opened the door at 1.00 p.m. on Monday, slowly and cautiously, by pushing it […]’ When we ask what logical form this expanded action sentence has, it is plausible to answer that it involves existential quantification over token actions, so that it is logically equivalent to something like this:

(∃a)(a was a door-opening and John was the agent of a and a occurred at 1.00 p.m. on Monday and a was slow and a was cautious and a was done by pushing …).

Taking this to be the logical form of our expanded action sentence, we can easily explain, for example, why it entails our original action sentence, ‘John opened the door’: it does so simply because a conjunction entails each of its conjuncts. However, if we then accept, in addition, W. V. Quine’s (1969) criterion of ontological commitment – encapsulated in his famous dictum ‘to be is to be the value of a variable’ – we may conclude that action sentences like these are implicitly committed to the existence of token actions, as the items quantified over by such sentences when their underlying logical form is made explicit (see chapter 6).

Of course, Quine’s criterion of ontological commitment is by no means uncontroversial and, in any case, even if it tells us that we are ontologically committed to token actions, it still doesn’t really tell us what these items are. The usual presumption, however, of those who follow this line of argument is that actions are events, even if not all events are actions: that is, they form a sub-class of events. This is because it seems natural to describe events in general, as well as actions in particular, as being individual occurrences that possess a particular time of occurrence. On this view, what is distinctive about actions is that they always have agents and also, perhaps – at least according to philosophers such as Davidson (1971) – that they are always intentional under some description of them. By contrast, it seems that there are many events, such as the explosion of a supernova in the Andromeda galaxy or the spontaneous decay of a radium atom, that have no agent and are not intentional under any description of them.

Suppose we agree, at least provisionally, that actions are events, although this has been disputed by some – for instance Kent Bach (1980). It then remains to be asked what events are. Two views on this issue are particularly dominant at present. One is Davidson’s own view, which is that events constitute a basic and irreducible ontological category of particulars, equally fundamental with that of physical objects (things such as John, or a radium atom). The other is Jaegwon Kim’s (1976) view that events are property exemplifications: more precisely, that an event is the exemplification of a property by an object at a time. On this latter view, each token event may be represented by an ordered triple of an object, a property, and a time, of the form 〈o, P, t〉. So for example John’s token action of opening the door, assuming it to be an event, may be represented by the ordered triple 〈John, door-opening, 1.00 p.m. on Monday〉. According to this view, events do not constitute a fundamental ontological category of particulars, since they may always be analyzed in terms of items which belong to other categories: the categories of objects, properties, and times. It might be objected to the Kimian view that it fails to register the fact that events are changes and conflates them with states – a state being a condition which does not involve change. In reply, however, it might be urged that the distinction between changes and states is, at best, superficial and sometimes difficult to adjudicate upon: for instance, is uniform motion in a straight line (inertial motion) a state of the moving object or a change in it?

How should we decide between the Davidsonian and the Kimian views of events, presuming that we should adopt one of them? The Kimian view might seem to be ontologically more extravagant because, while it analyzes events in terms of objects, properties and times, it still leaves us with at least these three basic ontological categories, whereas the Davidsonian view is apparently committed only to two: objects and events. On the other hand, Occam’s razor only enjoins us not to multiply entities (and, by implication, fundamental categories of entities) beyond necessity – and it may be argued that we need to include properties in our ontology in any case, for all sorts of explanatory purposes (for instance, to give adequate accounts of causation and causal laws).

Before leaving this issue, however, I want to revisit the question of whether actions really are a sub-class of events. In some cases this assumption seems unproblematic, but in others not. Suppose, for instance, that we attribute to John the action of having killed Mary by shooting her. Suppose also, to make matters interesting, that, although John shot Mary on Monday, she did not die until Wednesday, by which time John had already committed suicide in an act of remorse, say on Tuesday. If John’s action of killing Mary was an event, then what was its time of occurrence? If we say that it occurred on Monday, then we are implying, counterintuitively, that John killed Mary two days before she died. On the other hand, if we say that it occurred on Wednesday, when Mary died, we are implying, equally counterintuitively, that John killed Mary a day after he himself died. The source of the difficulty might be traced to this: intuitively, for John to kill Mary is for John to cause Mary’s death, so that in this kind of case an action is an agent’s causing of an event. The event which is caused – in this case, Mary’s death – may quite unproblematically have a time of occurrence (in this case, it was on Wednesday). But what about the causing: does that plausibly have a time of occurrence? Take another example, which does not involve agents, but simply the causing of one event by another: the case of an explosion causing the collapse of a bridge. The explosion has a time of occurrence, as does the collapse of the bridge (even if these events are, each of them, spread out over a period of time, rather than being momentary). But does the explosion’s causing the collapse have a time of occurrence? Indeed, is it an event, in addition to the explosion and the collapse themselves? It is not so clear, I suggest, that the correct answer to either of these questions is ‘Yes.’ If causings, quite generally, are not events and at least some actions are causings, then not all actions are events, even if some are. We might have to conclude, on this basis, that actions don’t constitute a unified category of entities at all – not even a sub-category of some other category.

What Are the Identity Conditions of Actions?

The foregoing discussion feeds directly into another important ontological question concerning actions that was raised at the beginning of this chapter. How are actions individuated and – if this is a different question – what are their identity conditions? Since the issue of action individuation is a leading theme of chapter 2, I can afford to be fairly brief here as far as this question is concerned. The word ‘individuate’ has two importantly different senses: a cognitive one and a metaphysical one. In the cognitive sense, individuation is the singling out of some entity in thought. In the metaphysical sense, it is a mind-independent determination relation between entities. It is in the former sense, for instance, that the police witness may be said to have individuated the perpetrator of the crime at an identity parade. It is in the latter sense, however, that we may say, for example, that a set is individuated by its members: for it is the members of a set, and they alone, that determine which set it is – they fix its identity. Since we are concerned in this chapter only with the ontology of action, we shall consider here the individuation of actions only in the metaphysical sense of the word ‘individuation.’

What, then – if anything – determines which action a given action is (assuming that we are still talking here exclusively about token actions)? On the Kimian view of events and actions, the answer seems straightforward enough: a certain object, property, and time always jointly determine this, for an action just is the exemplification of a certain property by a certain object at a certain time. This also provides us, immediately, with a criterion of identity for token actions, in the following form: If a and b are token actions, then a is identical with b if and only if a and b are exemplifications of the same property by the same object at the same time. However, on the Davidsonian view, no such easy answer is forthcoming. Davidson himself (1969) originally proposed a causal criterion of identity for events – and hence for actions – along these lines: If e and f are token events (or actions), then e is identical with f if and only if e and f have the same causes and effects. But it was soon pointed out that this criterion seems problematic, because it appears to be implicitly circular, at least on the assumption that all causation is causation by and of events. For then to say that e and f have the same causes and effects is just to say that the same events cause e and f and the same events are caused by e and f. Yet the criterion is supposed to tell us under what conditions events are the same or different, and so it shouldn’t just presume that, where the causes and effects of e and f are concerned, this can be regarded as being already settled.

Even if this problem can be overcome, the Davidsonian criterion of identity for events and actions raises another contentious issue: namely whether a criterion of identity for events tells us how events are individuated, in the metaphysical sense of ‘individuate.’ It is not clear that it necessarily does so. For an account of what individuates an entity x is supposed to tell us what determines which entity of its type x is: and it should presumably tell us this even with regard to counterfactual circumstances in which x may be supposed to exist, not just with regard to its actual circumst...