An Anscombean Perspective on Habitual Action

Abstract

Much of the time, human beings seem to rely on habits. Habits are learned behaviours directly elicited by context cues, and insensitive to short-term changes in goals: therefore they are sometimes irrational. But even where habitual responses are rational (contributing to current goal fulfillment), it can seem as if they are nevertheless not done for reasons. For, on a common understanding of habitual behaviour, agents’ intentions do not play any role in the coming about of such responses. This paper discusses under what conditions we can say that habitual responses are, after all, done for reasons. We show how the idea that habitual behaviour cannot be understood as ‘acting for reasons’ stems from a widely but often implicitly held theoretical framework: the causal theory of action. We then propose an alternative, Anscombean understanding of intentional action, which can account for habitual responses being done for reasons.

1 Introduction

In our everyday lives, we usually do not need much deliberation in order to get things done. For the most part, we make coffee, commute to work, interact with others, and prepare meals without any need for explicit, step-by-step reasoning regarding what to do next. In other words: for most of our everyday behaviour, we rely on our habits. In psychology, habits are usually defined as “learned automatic responses with specific features” (Wood and Rünger 2016). They are directly elicited by context cues, and they are insensitive to short-term changes in goals. To give an example of the first feature: once you have got into the habit of going for a run every morning at 9 AM, you will unthinkingly reach for your running shoes as soon as the clock shows it to be 9 AM. The environmental stimulus is all that is needed to start off the habitual response. The feature of insensitivity manifests itself in the fact that you might automatically reach for your running shoes at the usual time even if you for example seriously injured your knee the day before. But even though habitual behaviours thus seem to be ‘unreflective’ in some sense, they often seem to be intentional in the sense that many habits clearly bear an important relation to our long-term goals and plans. This paradox raises the question that will be the topic of this paper: are habitual behaviours guided by goals and intentions, and under what conditions does such guidance allow us to say that our habitual responses are done for reasons? This comes down to the question: can habitual behaviours be actions?¹

Before we can address this question systematically, we need to clarify how we define habits. Even though the psychological definition of habitual behaviour is widely accepted, it has been criticized from various perspectives. Especially the idea that habits are inflexible responses has encountered serious criticism. Quite recently, Turner and Cacciatori (2016) have argued that psychology almost exclusively focuses on so-called ‘automatic habits’ which are indeed characterized by inflexibility. However, they argue that three additional forms of habitual behaviour can be identified. First of all, so-called skillful habits (such as sports routines) are actually highly flexible with regard to variations in context. Think of basketball players, who must continuously tune their movements to the changing situation on the field.²

Secondly, some habits are contested in that agents actively exert self-control in order to control or suppress the habitual behaviour (Adriaanse et al. 2010; Quinn et al. 2010). Even if the habitual response itself is still automatic and inflexible (as pointed out in (Orbell and Verplanken 2015), it suggests that habits are not in principle insusceptible to self-control and deliberation. And thirdly, sometimes habits even seem to be both flexible and what Turner and Cacciatori (2016) call infused by deliberation, in that the agent can employ deliberative capacities in order to actively support or guide the performance of habitual behaviour in different situations.

Turner and Cacciatori argue that acknowledging the role of skillful, contested, and infused habits is crucial for understanding habitual behaviours of individuals and groups. More in general, their account can be seen as a call for a widening of the traditional psychological conception of habit. In order to do justice to these critical discussions, we here adopt a wider notion of habits as behavioural responses that do not depend on explicit deliberation, working memory or effort for their manifestation, thus leaving room for non-automatic habits.

Habitual behaviour is generally seen as the default mode of responding (Aarts and Dijksterhuis 2000; Arpaly 2000; Bargh and Chartrand 1999; Wood et al. 2014): we rely on our habits unless the situation requires us to overrule them by explicit deliberation and self-control. This interpretation builds on familiar dual-process models of cognition of the ‘default-interventionist’ kind, according to which cognition generally operates via Type 1 processes (that are automatic in the sense of not requiring working memory) unless the agent responds to situational demands by initiating cognitively more ‘expensive’ Type 2 processing (demanding working memory and some form of cognitive decoupling, (Evans and Stanovich 2013)). Even though such reliance on habitual responses is generally taken to be highly adaptive (Cosmides and Tooby 1996; Gigerenzer and Gaissmaier 2011; Oaksford and Chater 2007), the insensitivity of habitual responses to current changes in goals and expected outcome still raises important questions concerning human rationality. Even though habits themselves might be adaptive, changes in situation can make concrete manifestations of habits irrational. An often-used example is the story of a group of Soviet soldiers on a secret mission in Cuba who, upon landing, unthinkingly marched away in perfect formation (Allison 1971; Turner and Cacciatori 2016). The soldiers’ response is instrumentally irrational because even if the response contributed to goal-fulfillment at their military home base (being a good soldier requires marching in formation), it does not contribute to —in fact even undermines— the fulfillment of their current goal of remaining undercover abroad. Such instrumental irrationality can occur because habitual responses are not elicited by the agent’s current goal states or intentions, but by environmental cues (Ji and Wood 2007; Neal et al. 2012; Neal et al. 2011). If the latter is the defining characteristic of habitual action, this may seem to suggest that even rational habitual responses (thus responses that contribute to current goal fulfillment) do not seem to be done for reasons: after all, agents’ current goals or intentions do not seem to play any relevant role in the coming about of such responses (Makowski 2017).

Such worries bring us to the main question of this paper: under what conditions can we say that our habitual responses are done for reasons? In the next section, we will first set the stage by discussing how habitual responses are usually understood as forms of intentional action. We will show how a widely but often implicitly held theoretical framework (the causal theory of action) seems to lead to the conclusion that habitual action, even though it prima facie appears to be intentional, cannot be understood as ‘acting for reasons’. In the next section we will then proceed to develop an alternative understanding of intentional action, and show how such a perspective can account for habitual responses being done for reasons.

2 Habitual Behaviour: Done for Reasons?

Habitual behaviour is generally taken to be intentional or goal-directed in at least an indirect sense. After all, we usually develop a certain habit because we are trying to attain a certain goal (Arpaly 2000; Anderson 1982; Wood and Neal 2007), and manifestations of that habit can thus be seen as being directed towards the achievement of that goal. For example, the habit of reaching for one’s running shoes at a certain time of the day has been acquired because for the person in question, running is a goal. Over time and repeated exercise, the idea is, this goal-directed response becomes so automated that at some point it is directly elicited by the relevant cues without any reliance on working memory, explicit deliberation or control. Whereas some argue that habitual behaviour should be understood as a form of automatic goal pursuit (Arpaly 2000; Wood and Neal 2007), others object that because habitual responses are insensitive to current goals, habitual behaviour is only derivatively goal-directed (Wood and Rünger 2016). But however conceived, the connection between habitual behaviour and goals distinguishes it from other forms of automatic processes such as preconscious perceptual processing, which is not aimed towards a goal in any sense (Bargh and Chartrand 1999, pp. 463–4).

Whereas most psychologists and philosophers seem to agree that habitual responses are goal-directed in the sense discussed above, it is generally thought that such responses are not brought about by intentions (Railton 2009). As already mentioned, the psychological literature suggests that strongly habitual responses are heavily ‘underpredicted’ by motivational states such as goals and intentions, as they are directly elicited by environmental cues (Ji and Wood 2007; Neal et al. 2012, 2011). This has led to the philosophical worry that it might not be possible to say that such responses are done for reasons (Brownstein 2014; Makowski 2017). Brownstein (2014) suggests that it is in fact paradigmatic for habitual behaviour that agents cannot explain, or can only falsely explain, why they do what they do: the reasons they provide should be seen as post hoc rationalizations, as they do not track the causal processes that led to the behavioural response.

However, the conclusion that habitual behaviours are not done for reasons seems prima facie puzzling. As Turner & Cacciatori (2016) have argued (see the introduction), it seems that much of our everyday behavior that is not preceded by explicit deliberation is nevertheless ‘infused’ with rationality: not every case of habitual behaviour seems to be a case of automatic behaviour, in the sense of being fully “mindless” or completely blind to our reasons. The case of our agent reaching for her running shoes, for example, still intuitively seems like a case of intentional action: the response is different from a mere reflex movement. In the remainder of this section we will argue that this difficulty about understanding habits as ‘acting for reasons’ arises from the familiar assumption that behaviour is done for reasons only in so far as it is caused by reasons (in other words, the assumption of causalism, or the causal theory of action).

Building on a Humean tradition, causal accounts of action have been developed by, e.g., Donald Davidson (1963) and Michael Smith (1994). In psychology, the approach has become most widely known in the form of Michael Bratman’s belief–desire–intention model (Bratman 1987; Bratman et al. 1988). In the BDI-model, agents weigh different reasons for action on the basis of their goals: the reason that comes out strongest subsequently causes the formation of a corresponding intention, and this intention in turn causes the corresponding action.³ This approach to action explanation has become widely popular in both cognitive science and various approaches to artificial intelligence. It shares important assumptions with traditional functionalism, which also builds on the idea that mental states should be defined in terms of their causal role in the mental architecture (Armstrong 1981; Fodor 1975). It is easy to observe that almost all psychological literature on the role of goals and intentions in habitual behaviour starts from such a causalist/functionalist assumption: if goals and intentions have a role to play in habitual behaviour, their role should be a causal one. For example, Neal et al. (2012) distinguish between the ‘inferred’ or ‘perceived’ role of goals on behaviour and their ‘actual’ role, which is their effective causal contribution, measured in terms of cognitive associations (2012, p. 493, see also Wood & Rünger 2016, p11.15).

Now, how does such a causalist framework raise doubts regarding the question whether habitual responding is ‘acting for reasons’? The answer is simple. Firstly, dual-process approaches to cognition suggest that behaviour is caused by reasons in so far as it is caused by a process of deliberation: conscious consideration and weighing of reasons (Chan 1995; Korsgaard 2009). Such deliberation is characterized as typical Type 2 processing: it is time-consuming, effortful, and relies on working memory and cognitive control (Arpaly and Schroeder 2012; Makowski 2017). However, as described in the introduction, habitual responses result from Type 1 processing that precisely does not rely on working memory and cognitive control (Wood et al. 2014; Wood and Rünger 2016). From the perspective of a causalist approach to action explanation, this suggests that habitual behaviour is precisely not the kind of behaviour that can be explained by referring to any deliberative process involving reasons as causal factors.

So when looked at from the perspective of dual-process models of cognition, a causalist approach to action explanation seems to suggest that ‘acting for reasons’ stands opposed to ‘responding habitually’. Acting for reasons is defined as activity in which a deliberative process plays a causal role, and as such is caused by Type 2 processes that are categorically different from the kind of processes (Type 1) that instigate habitual responses.

3 Prior Attempts to Create Room for Habitual Actions

Within the paradigm inspired by the causal theory of action, there have been various attempts by authors to show how habitual behaviour could nevertheless be seen as action done for reasons. Recent examples include Railton (2009), Arpaly and Schroeder (2012), and Velleman (2008).⁴ In this section we will discuss their proposals and show that they lead to problems stemming from their more or less tacit assumption of the causalist paradigm.

These authors agree with us that behaviours we are typically inclined to call habitual are in an important sense different from mere mindless behavior. As Railton (2009, p. 98) puts it: “Far from being merely swept along by a causal chain or robotically enacting a habit or routine, the jazz improviser and skilled basketball player, for example, are exercising agency in a pure form, fully deploying their cognitive and creative skills.” That is: Railton agrees that there is a sense in which these behaviours are both habitual and done for reasons. But he (and the other authors mentioned above) attempt to explain this from within the causalist paradigm. The strategy they employ to reconcile the notion of habituality with that of acting for a reason is to argue that acting for reasons does not, after all, require conscious deliberation. They suggest that what it is to act for a reason just is for one’s action to be caused by certain mental states and processes. However, they argue, the relevant states and processes may simply be unconscious, and need not figure in any deliberation of the agent’s. What is essential to behaviour being done for a reason is just that there is a rationalizing relation between the relevant mental state (say, the goal to go running) and the behaviour in question (say, reaching for one’s running shoes), and that the mental state is the (unconscious) cause of the behaviour. For example, Arpaly argues that an agent can be said to act for a reason in so far as the mechanism causing the behaviour is responsive to reasons in a very minimal sense, namely in the sense that it is “the result or the embodiment of an awareness, inaccessible at the moment to his deliberation” of what is good about the behaviour (Arpaly 2000, p. 503). And Railton argues that, in order for it to be possible that automatic or habitual responses of the kind described above (the saxophone player’s solo, the skilled driver’s maneuvers, or our agent’s habitual reaching for her running shoes) are done for a reason, “there must exist non-deliberative causal psychic processes ‘of the right kind’ to be aptly responsive to a given consideration as such, and aptly expressive of one’s identity or values, even in the face of competing interests” (Railton 2009, p. 104).⁵

Although we believe there is something right about these attempts to do justice to the phenomenon of habitual action, we believe that there are important difficulties with the proposed causalist solution in terms of unconscious causation. First, there is the initial problem that it is unclear what epistemic right the causalist has to postulate the existence of unconscious beliefs and intentions (or other mental states or processes). Given that they are unconscious, how are we to know that they are present, exactly in those problematic cases of habitual or automatic behaviour that we intuitively still want to describe as intentional? Given the claim from empirical psychology that much of our everyday behavior might actually not be done for reasons, that may seem like a mere ad hoc move.

But even if we grant the causalist the hypothesis that there are such unconscious mechanisms at play precisely in those cases that we intuitively want to describe as happening for reasons, we argue, there remains a deeper problem with this strategy to reconcile causalism with the idea that habitual behaviours happen for reasons. As we have just seen, Railton interestingly included the clause that the unconscious processes that explain habitual behaviour must be ‘of the right kind’. What does this mean? Clearly, the clause is intended to exclude the possibility of so-called deviant causal chains. For, as Railton insists, the habitual responses under consideration “are clearly done for reasons, and, moreover, for reasons as such (rather than, say, through a deviant causal path)”. We agree that what is needed to understand the sense in which the habitual responses under consideration are not ‘merely mindless’ is how they can be done for reasons ‘as such’. That is, it is not enough that there is some explanatory link between the agent’s reasons (i.e., her desires or ‘pro-attitudes’, broadly speaking) and her action: if the agent’s reasons would cause her behaviour merely accidentally, then we would still not be able to view the behaviour as rational in the required sense. And such an accidental link between mental states and behaviour is precisely what seems to be present in the famous cases of deviant causal chains, which are frequently discussed as a threat to the entire causal theory of action. It will be helpful to consider the famous example of a deviant causal chain developed by Davidson himself:

A climber might want to rid himself of the weight and danger of holding another man on a rope, and he might know that by loosening his hold on the rope he could rid himself of the weight and danger. This belief and want might so unnerve him as to cause him to loosen his hold. (Davidson 1973, p. 79).

There is, of course a large and still growing body of literature on the question how much of a problem the possibility of deviance is for the causalist paradigm, and on whether it is possible to define the required right causal connection between mental states and behaviour. Here, however, our aim is not to answer the question whether there is, ultimately, a way out of the problem of deviant causal chains. What we will point out below is merely that there is a tension between the strategies that causalists adopt for dealing with the problem of habitual actions on the one hand, and the general strategies available for circumventing deviant causal chains on the other. For the latter strategies appeal to precisely those features that Railton, Arpaly, and others insist must be absent in the case of habitual behaviour: the agent’s consciousness and deliberation.

To see this, consider Davidson’s own initial diagnosis of what is missing in the climber case. He observes that in the ‘right’ cases, the attitudes that constitute an agent’s reasons cause the action “through a course of practical reasoning”, whereas in e.g. the climber scenario, this is precisely not the case. Davidson is famously sceptical about the possibility of giving a general account of the conditions under which reason-states cause the action in the right way, i.e. “through a course of practical reasoning”. Other defenders of the causal theory of action are more optimistic about this, offering at least partial analyses of when a causal chain involving reason-states constitutes a piece of practical reasoning: for instance, by insisting that the agent’s attitudes concern not only the desired end result of his action, but also the more or less precise way in which this result is to be achieved (i.e., the action must be caused by a state representing the agent’s plan for action)—that is, by insisting that the agent represents his means as such. Others argue that the attitudes that cause the action must somehow be represented or endorsed by the agent as reasons for action (Schlosser 2012): the agent’s reasons must, e.g., include a reference to themselves, such that the agent is moved by a desire not just to do A, but to do A because of this very desire.⁶

But however the causalist wishes to spell out the relevant details, it seems clear that these definitions of the “right” kind of causal chain, understood as a definition of when causation by reason-states constitutes “a course of practical reasoning”, will not be met by habitual behaviour as e.g. Railton construes it. For Railton stipulates that agents responding habitually are not conscious of the reasons’ causal efficacy, that they do not engage in practical deliberation, and that they do not choose or endorse their reasons as reasons (Railton 2009 p. 102).⁷ In the face of this, how can it still be maintained that in such cases their behaviour is explained by reasons ‘as such’?

Authors like Railton may claim that it is simply evident that habitual behaviour such as the musician’s sax-playing, or our agent’s reaching for her running shoes, happen for reasons. Now, as we have explained, we agree with this assessment. But if we respect the distinction between behaviour merely being caused by a reason, and its happening for a reason ‘as such’, or for a reason qua reason, it is not enough to be told that habitual behaviours are produced by unconscious mechanisms. What needs to be shown is that such mechanisms, that are said to produce habitual behaviours entirely behind the agent’s back, nevertheless constitute causal chains of the ‘right’ kind. And as the brief diagnosis of cases of deviance provided above indicates, there is reason to be sceptical about that claim: consciousness and deliberation, in some sense, seem to be required to rule out that the connection between an agent’s reasons and her behaviour is merely accidental.

We take this to be an important lesson to be learned from the causalist attempt to account for automaticity. While we agree with e.g. Railton’s assessment of the intuitively habitual or automatic responses under discussion here as not merely mindless behaviour, but as intentional, it seems that we cannot explain this on the assumption that the efficacy of the agent’s reasons happens entirely outside the purview of the agent’s consciousness. What is needed is to explain in which sense agents do have their ends or reasons ‘before their minds’, even when they respond habitually—while respecting the fact that such behaviour is precisely characterised by the absence of explicit prior deliberation. We suggest that to do justice to this point and to understand habitual action, it is necessary to take a step back from the causalist framework.

4 Acting for Reasons and the Question ‘Why’

We have argued that there is serious pressure on the commonsense idea that at least some of our habitual behaviours are intentional and happen for reasons. This pressure stems, on the one hand, from empirical psychology—which suggests that an agent’s prior attitudes and her prior deliberation are not relevant to explaining habitual responses—and on the other, from the causalist paradigm in the philosophy of action—which, pace e.g. Railton, seems wedded to making the presence of such prior deliberation and attitudes a necessary condition of behaviour happening for reasons (or for reasons ‘as such’). In this section, we wish to suggest that it is possible to rescue the commonsense insight that habitual or automatic behaviour happens for reasons (and are thus not completely mindless) by showing that there is another way to understand what it means to ‘act for reasons’—an approach which does not rely on the causalist/functionalist approach to action explanation. In particular, the approach we have in mind⁸ rejects the following two assumptions, which, as we can now see, generate the difficulty concerning the possibility of habitual action:

• A Practical deliberation is an event or process that necessarily occurs prior to the action which is explained by it.

• B An agent’s consciousness or awareness of her reasons for acting must be an occurrent episode, in the sense of being in the forefront of her mind, and thus requiring a certain form of attention or effort.

The alternative tradition in the philosophy of action which rejects these assumptions is primarily represented by Elizabeth Anscombe (1957), but has in recent times received much renewed attention, and has been further developed by, e.g., Moran (2004), Thompson (2008), Marcus (2012) and Lavin (2013). Roughly, the idea behind these accounts is that intentional action is action of which agents know that they are doing it and why: they can answer the so-called “why-question”, or the “Anscombean question”, as Railton (2009) calls it. However, and crucially, answers to that question should not be seen as reports on any preceding internal processes (let alone processes involving deliberation over reasons): instead, they should be understood as manifestations of a form of practical knowledge, or insight into the goal-directed structure of one’s actions. As we will argue, this offers a different perspective on habitual action: in so far as agents have practical knowledge of the goal-directed structure of their habitual behaviour, such behaviour can be both seen as intentional and as ‘done for reasons’.

The guiding thought behind Anscombe’s account is that intentional action is action to which a certain sense of the question “Why?” is applicable. Intentional action, Anscombe’s thought is, is the kind of behavior for which it makes sense to expect the agent to be able to give an explanation. And the relevant sense of the question “Why?” is, of course, the sense in which it is a demand for the agent’s reason for action. Anscombe, and philosophers of action who follow her, argue that the required elucidiation of the relevant sense of the question “Why?” can be found by seeing that agents have a certain kind of knowledge of the answer (when they are indeed acting intentionally). This kind of knowledge can be negatively characterized as “knowledge without observation”. Positively, we can say that such knowledge is practical knowledge. We will now briefly explain both points by examining four hallmarks of such agential knowledge.

Anscombe argues that the question “Why are you doing A?” does not have application when the agent does not know that she is doing A.⁹ For imagine an agent who, being asked why she is ringing the doorbell, answers: “’Good heavens! I didn’t know I was ringing it!’” (p. 51)—for example, the agent was unwittingly leaning against the doorbell. In such a case, the demand for a reason is obviously ill posed, and so the behaviour is not intentional. But, Anscombe continues, the question is equally refused application if the agent does know that she is doing A, but only because she observes herself to be doing it. For example, suppose the agent from our example notices that she is, in fact, ringing the doorbell by leaning against it. Then again, she is obviously not in a position to answer the question why she is doing that, in the relevant sense. As Wittgenstein says, the mark of intentional action is the absence of surprise¹⁰: when we do something intentionally, we do not need to observe ourselves doing it in order to know that we are doing it. And in so far as we do observe something we do intentionally, this will not surprise us: when one intends to raise one’s arm, the observation of one’s arm rising does not come as something unexpected. Furthermore, the same would hold if our agent would know that she is ringing the doorbell, but only because she reasons “Someone is ringing the doorbell; I am the only one near the doorbell; therefore I must be ringing it”. So we can say that, in order for the question “Why are you doing A?” to be applicable, and so in order for A to be an intentional action, the agent must know that she is doing A without observation (and without inference: she does not conclude that she is acting intentionally based on, say, her noticing that she has a desire to do A and finding that she is in appropriate conditions to do A).

Second, and relatedly, the question “Why?” is refused application if the agent fails to realize that it is she herself that is doing A. That is, it is not enough that she knows that, say, someone is doing A: for example, if she thinks “the one who is ringing the doorbell is waking the baby”, then neither ringing the bell nor waking the baby will be an intentional action of hers. Rather, her knowledge must be first-personal: i.e., it must be knowledge of what she would express, by using a first-person pronoun, as “I am doing A”. According to Anscombe’s account, intentional action must be thus be understood as essentially self-conscious: an agent is, in a sense to be understood, not at a distance from her action, so that she does not need to observe or infer that she herself is doing it. That is why, e.g., Davidson’s climber (see Sect. 2) is not acting intentionally: for him, the awareness that he is letting go of the rope because he is nervous is at best a realization or discovery, to which he is a passive bystander.

Third, it is important to note that an agent’s knowledge of what she doing is paradigmatically not just of the fact that she is doing A, but rather of the fact that she is doing A because she is doing B, which is itself an intentional action.¹¹ So the agent’s knowledge will be, for instance, “I’m chopping onions because I’m making risotto”. This is why the kind of knowledge without observation that we have of what we are doing intentionally can be called practical knowledge: in the central case, it is knowledge of the means-ends structure of one’s action.¹² Consider this chain of questions and answers to the Anscombean question:

• “Why are you grabbing the onion?” (A) – “I’m cutting some onions” (B).

• “Why are you cutting some onions?” (B) – “I’m making risotto” (C).

• “Why are you making risotto?” (C) – “I’m preparing a Christmas dinner” (D).

Practical knowledge, as Anscombe argues, comprises this whole structure: the agent knows that she is doing A because she is doing B, which she is doing because of C, which in turn she is doing because she is doing D. Equally, she therefore knows that she is doing A because she is doing D.

Finally, this means that practical knowledge is present throughout the agent’s performance of her action. While she is cutting the onions, she knows that she is doing this because she is making risotto. And if she then grabs an extra onion from the kitchen cupboard, she will equally know that she is doing this (grabbing the onion) because she is cutting onions in order to make a risotto. On this view, there thus does not need to be a prior process of thinking (“I need to grab an onion for the risotto”) before the agent initiates the action. What makes the action a goal-directed performance is that the agent knows, as she is extending her arms to the cupboard, that this movement is part of a larger intentional action of hers.¹³

On the Anscombean view of action, what it is to perform an intentional action (doing A) is thus to understand what one is doing as contributing to one’s aims (“I am doing A because I am doing B”). The concept of an intentional action, then, is essentially the concept of a kind of behavior that makes sense to the agent as her action. An intentional action is action that is, therefore, partially constituted by the agent’s point of view, or her own take on what she is doing. According to the Anscombean view, this, and not an appeal to special mental causes, is what distinguishes between intentional action and mere behaviour.¹⁴

Naturally, this Anscombean account of practical knowledge and intentional action requires much further development, which we cannot undertake in this paper. But because it is so central to the Anscombean approach that an intentional action is understood by the agent herself, and therefore makes sense to her as contributing to her goals, it provides a fruitful way of thinking about precisely the cases of habitual or automatic responses that we have been considering. For, e.g., in cases such as that of our agent who habitually reaches for her running shoes, the reason why we intuitively want to classify this as an intentional action (and not a completely mindless happening) is precisely that doing what she does makes sense to the agent (Makowski 2017). And from the perspective we have been presenting in this chapter, that may be all that is needed for the behaviour in question to be intentional: the agent has practical knowledge that she is reaching for her shoes in order to go running.

Moreover, the Anscombean suggests that practical reasoning just is knowing the relation between means and ends in this way. This comes out in the fact that the chain of “Why?” questions and answers in the A-D series above can equally be traversed in the opposite direction: rather than beginning by asking the agent “Why are you doing A?”, we can begin with asking “How are you doing D?”, to which she can reply “by doing C”—and so on. In knowing that she is doing A because she is doing B, the agent simultaneously knows that B is her means to doing A. Practical reasoning and practical knowledge are thus, as it were, two sides of the same coin. On this view, practical reasoning is thus not an occurrent process which takes place before the action begins. Rather, practical reasoning is an awareness of what constitutes means to one’s ends that, as we have seen, is constitutive of the action in question. As such, it is an awareness that is manifested in, and thus lasts throughout, the agent’s performance of the action.

But doesn’t the Anscombean view overestimate the extent to which agents are able to give true explanations of their intentional actions? The empirical literature seems to abound with examples which seem to suggest that an agent’s answer to the question “Why?” is often just a post hoc rationalization, or a confabulation. This may especially seem to be so in the case of habitual behaviour. Brownstein (2014), for instance, suggests that empirical evidence supports this view with regard to at least highly skilled actions in sports (basketball, in the case at issue):

Confabulation is common when skilled agents are asked to report the techniques they use “on the field.” For instance […] players who successfully catch objects falling at an accelerating rate report that those objects are falling at a constant speed. These individuals are confabulating reasons - based in naïve physics - when they are asked why they are moving to the spot where the accelerating object is? (Brownstein 2014, p. 560).

Notice, however, that the kind of explanation Brownstein is here taking to be the player’s answer to the “Anscombean question”—namely, “the ball is falling at constant speed”—is not actually the kind of consideration that the Anscombean claims to fall under the scope of the agent’s practical knowledge. For it is obviously not the kind of thing that an agent could know without observation: it is not even a candidate for a reason for action, on the Anscombean view. So the fact that agents speculate or confabulate about the physical properties of the ball does not show that the behaviour is not intentional according to the Anscombean’s criteria. For that to be the case, it would have to be shown that the agent was not aware of the following fact: that she was moving to spot X on the pitch in order to catch the ball. Nothing in Brownstein’s example suggests that this is the case, and indeed, it seems obvious that the basketball player does know this as she is moving down the pitch. If we are thus attentive to what the Anscombean is claiming to the be the object of an agent’s practical knowledge—namely, the means-ends structure of her ongoing action—it will seem far less convincing that agents’ account of this structure could be, in general, mere confabulation. After all, on the Anscombean account, what an agent can truly be said to be doing intentionally depends on what she takes herself to be doing. Instead, what the confabulation literature shows is that agents provide many causal explanations of their actions that turn out to be faulty. So whereas Anscombe’s view leaves room for a lot of self-deception and failing self-insight in human agents—it does not leave room for agents being mistaken about the means-end structure of their own actions. This is because, on the Anscombean picture, action is defined in terms of this kind of practical knowledge.

This also shows how Anscombe’s view would respond to the well-known examples from empirical psychology in which it seems as if agents are actually unaware of the ‘true’ reasons for which they act, such as the famous Nisbett & Wilson pantyhose experiment (Nisbett and Wilson 1977), in which agents seem to confabulate reasons for picking one of a number of identical pantyhoses. The Anscombean response would be that even though there is certainly confabulation going on here, what is confabulated is not the means-end structure of the participants’ behaviour (in this case, their choosing one particular pantyhose). Their assignment was to choose the pantyhose they considered to be of the best quality, and this is what they did and what they said they did: they chose the pantyhose they considered to be of the best quality. What participants could not accurately report upon was the psychological factors that explained why they chose the one they did (given that they were all of identical quality). As Sandis (2015) aptly remarks: “What we are fabricating in such as case is not a tale about our agential reasons but one about the quality of the stockings” (p.270). The results of such experiments thus do not seem to stand in opposition to the Anscombean theory. It may be that the reasons cited by the agent are actually bad reasons (i.e., there actually are no differences between the objects of choice), but they are still her reasons, i.e. the reasons for which she picked the one over the other. The idea that ‘the real reason’ for choosing a certain pantyhose was an non-conscious preference for objects placed in a certain position (the psychological explanation) only makes sense within a causalist paradigm where ‘one’s reason’ is defined as the preference that caused one’s action.

The Anscombean proposal thus differs from the attempted solution to the problem of the intentionality of habitual responses offered by, e.g., Railton in Sect. 2 in the following way. Railton and others in the causalist framework are forced to postulate unconscious states of intention (or other pro-attitudes). And because they conceive of practical reasoning as an occurrent train of thought prior to the action, it follows immediately that these unconscious states cannot be part of practical reasoning. By contrast, the Anscombean can hold on to the idea that a form of consciousness—practical knowledge—is essential to acting. But such consciousness does not need to consist in the presence of some occurrent train of thought prior to action. Rather, it is knowledge of the point of the action, as the agent conceives of it. Part of what stands in the way of accepting this suggestion is, perhaps, an overly “intellectualist”¹⁵ conception of knowledge, as if knowing consists in occurrently having a thought before one’s mind. Brownstein (2014), perhaps, betrays this tendency when he writes that being able to answer the Anscombean question requires “thinking propositional thoughts about where your body should be or what it should be doing” (p. 558), and when he then complains that the effort of thinking such thoughts can only distract the agent from what she is doing.

Against this conception, we should note that we need not think that knowledge, in general, consists in occurrently having a proposition before one’s mind: everyone knows facts that they do not have occurrently before their minds. And in particular, we should note that when it comes to self-knowledge, such a picture of having a proposition occurrently before one’s mind seems mistaken: in general, one does not need to contemplate the fact that one has a certain belief in order to know that one has it.¹⁶ After all, when one thinks that p, one does not need to contemplate any further facts to know that “I think p”. If the Anscombean claim that knowledge of what one does intentionally is a form of self-knowledge can be developed and defended, then it seems that we could understand how many of our actions can be habitual, in the sense of not requiring any prior deliberation, while at the same time being more than mindless behavior.

We should emphasize that we do not deny that there may be cases of habitual behaviours of which an agent has no practical knowledge, and which are therefore unintentional. A case, frequently discussed in the empirical literature and already mentioned in the introduction, that comes to mind is that of the group of undercover soldies who upon arrival at their destination unthinkingly marched away in perfect formation. Arguably, they were not aware of marching in perfect formation—or at least, we suggest that whether or not this is a case of acting for reasons depends on whether they were able to answer the Anscombean question. However, it seems that many of the responses discussed in the literature as habitual (because there seems to be no occurrent process of prior deliberation) can be said to happen for reasons, given the Anscombean view. Automaticity, in the sense of lack of prior deliberation, need not imply lack of reasons on the agent’s part. Indeed, we suggest that we may usefully distinguish between habits that are compatible with, and indeed arguably require, the presence of practical knowledge on the agent’s part, and habits that do not. These may constitute different species of the genus “habit”, that require a different treatment in the philosophy of action, and perhaps in empirical psychology as well.¹⁷

5 A new look at self-ascription of reasons and intentions in empirical psychology

In the introduction we discussed Turner and Cacciatori’s (2016) suggestion that most forms of habitual behaviour are much more flexible (in the sense of their being susceptible to goals and reasons) than is often held in the psychological literature. But as long as one holds on to the idea that reasons are causal processes that must take place prior to an action, the fact that habits are directly elicited by the environment seemed to exclude such flexibility. However, when we now look at Turner and Cacciatori’s proposal from an Anscombean perspective, we can see that it allows for precisely the kind of flexibility Turner and Cacciatori suggest. The fact that habitual responses are initiated on the basis of environmental cues does leave ample room for practical knowledge: for the agents involved to know what they are doing and why.

To clarify: from an Anscombean perspective, many of the empirical results we have discussed in Sect. 1 no longer seem threatening to the idea that habitual behaviour can be done for reasons. For, as we have seen, what seemed to threaten that idea was the fact that the agent’s reported intentions prior to the response underpredicted what the agent would do. But on the Anscombean proposal, that need not undermine the intentionality of the habitual response: even if the agent has expressed the prior intention to quit running because of an injury, what matters is that she is able to answer the question “Why?” during the performance of her action: she should be able to answer the question: why are you putting on your running shoes? The point here is that in such cases, the habitual response of putting on one’s running shoes is still intentional in the sense of it being done for the sake of going running. And in so far as the agent has practical knowledge, and thus can answer the question “why do you put on your running shoes?” by answering “I want to go running”, the habitual behaviour is clearly done for a reason.

If that answer contradicts a previously expressed intention, that is in itself an interesting result: maybe this indicates, for example, that agents frequently change their minds about what to do. It can also indicate (as the example hints at) that we are only reminded of our prior intentions when someone starts to question our actions. But however we want to explain that fact, it need not undermine the idea that the habitual response of putting on one’s shoes was performed for reasons, as long as the agent was able to answer the Anscombean question while she was engaged in it.

Also, the Anscombean account of ‘acting for reasons’ sheds new light on the results of self-reports concerning the role of goals and intentions in habitual behaviour. For example, Ji and Wood (2007) show that if people have strong habits, their behaviour is not best predicted by the strength or certainty of the intentions expressed, but by past behaviour. Based on such findings, Ji and Wood suggest that for people with strong habits, intentions might be epiphenomena (Ji and Wood 2007 p. 273). This confirms that they assume that in so far as the previously expressed intentions do not predict the relevant behaviour, they do not cause it, and therefore, on their view, such behaviour is not intentional at all. Neal et al. (2012) even think that this discontinuity between the perceived and actual role of such states shows “the need to study mechanisms of actual habitual control using methods that do not rely on introspection” (p.497). We believe the implicit reliance on a causalist way of thinking in the psychology of habits misrepresents the role of goals and intentions in habitual behaviour, and also misrepresents the status of self-reports of such states (namely, as reports of internal causal processes). As an alternative, the Anscombean framework we have outlined above offers the possibility to analyze self-reports of goals and intentions, not as reports on internal states with a causal functional role, but as manifestations of practical knowledge of the intentional structure of one’s behaviour. Expressing one’s intentions should be seen as engaging in a form of commitment, informed by one’s goals. And in so far as one’s subsequent actions correspond to such expressed intentions, this shows that these actions are done intentionally. This provides a much simpler explanation of the (to some) paradoxical fact that people with strong habits hold intentions with greater certainty, even though they “did not rely on these strongly held judgments to guide habitual behavior” (Ji and Wood 2007, p.273). All this is no longer paradoxical once one understands reports of goals and intentions as accounts of the ‘sense’ in our own behavioural patterns. We would say that people do rely on their intentions, but that guidance need not necessarily be interpreted as a causal chain from a prior mental state to an action: it can also be understood as practical knowledge shaping habitual action while it is unfolding.