Accounting for doxastic agency: Mental action and self-awareness

Abstract

Uncontroversially, individuals exercise agency in acting; can we say the same about believing? This paper argues that subjects do indeed exercise agency over their beliefs and provides an account by which this is possible. On my picture, self-awareness is fundamental to the nature of doxastic agency. Drawing on work in the philosophy of action, I argue that subjects exercise agency in performing mental actions that form and sustain their beliefs, where they are aware of these actions as part of reasoning and exercising agency over their beliefs.

Uncontroversially, individuals exercise agency in acting; can we say the same about believing? This question bears special importance for doxastic responsibility. If subjects lack agency over their beliefs, then it seems unfair to criticise someone (as we surely want to) for, say, believing that women are inferior to men. Yet formulating an account of this agency is difficult, especially because we cannot model it directly on intentional action – subjects cannot simply choose their beliefs. This paper supports the idea that subjects do indeed exercise agency over their beliefs by providing an account of the way in which this would be possible. On my picture, self-awareness is fundamental to the role played by mental action in doxastic agency. Drawing on Anscombe’s work in the philosophy of action, I argue that:

A subject exercises agency over her belief that p in being disposed to perform a range of mental actions in reasoning which form and/or sustain the belief, where she is aware of these actions as part of reasoning and exercising agency over her belief that p.

The way in which subjects view what they are doing helps make these actions the right sort to furnish them with agency over their beliefs. As such, my claim is metaphysical – self-awareness plays a constitutive role in doxastic agency.

This paper proceeds as follows. Sect. 1 introduces the notion of doxastic agency and the proposal’s importance and prima facie plausibility, but sect. 2 puts the thesis into question – it’s not clear that the most promising account of doxastic agency suffices as it stands. I seek to rescue the prospects of doxastic agency in sect. 3, though, by formulating the self-awareness account. Sect. 4 further supports this proposal in response to possible objections.

1 Introducing doxastic agency

1.1 The central claim and its appeal

This paper concerns the thesis that we possess doxastic agency – but what would this amount to and why is it initially appealing?

To answer these questions, it’s worth considering subjects’ agency, or lack of it, in the cases of action and pain.

ACTION: Julia arrives late to an important meeting. We may well criticise her: ‘you shouldn’t be so late!’. We might retract our blame if she has a good excuse, but nevertheless, arriving late is the sort of thing which we could blame Julia for.

PAIN: Beth has a headache and has to cancel a meeting. Her colleagues wouldn’t criticise or blame her for her headache, or say: ‘you shouldn’t have a headache at such an important time’. They might blame her if she didn’t do all she could to get rid of the pain, like taking paracetamol. They might also criticise her for not holding the meeting despite being in pain. But they wouldn’t criticise her for being in pain as such.

Prima facie, it is only fair to hold people accountable for things they have control over. We hold subjects accountable for their actions but not pains. It seems therefore that subjects exercise agency in performing physical actions but not in experiencing pain.

In this regard, belief resembles action. Consider the following.

BELIEF: Tom believes that the meeting will start at noon. Suppose it turns out that the meeting starts at 11am instead, or that the meeting does start at noon but it was a lucky guess. We would criticise him in these cases: he shouldn’t believe false things, and shouldn’t believe propositions willy-nilly.

At first sight, Tom exercises a form of agency over his beliefs, as do we all. Tom’s belief and Beth’s pain importantly differ in that we can hold Tom accountable for his beliefs – we can praise and blame him. Since we can hold subjects accountable for their beliefs and since, plausibly, we can only hold subjects accountable for things they can exercise agency over, it looks like subjects exercise agency in holding beliefs.

At this point I should emphasise three points concerning doxastic agency and how I understand the term. (1) There is no question about whether subjects have some form of indirect control over their beliefs.¹ For example, Tom might ask a hypnotist or a therapist to make it so that he believes that the meeting starts at 11am, or leave himself reminders. Tom could similarly also make it so that he continues to have this belief, such as by visiting the hypnotist on a daily basis and deliberately refraining from reading disconfirming sources. But here Tom only seems to exercise agency in getting himself to believe that p or to only exercise agency in visiting the hypnotist – not in holding the belief itself. It’s the same sort of control Tom could exert over Beth’s belief by taking her to a hypnotist or reminding her of facts to make them salient. This is like how we can get ourselves to be in pain by hitting our head or drinking too much. And we can even make it such that we continue to be in pain once in that state, e.g., by prodding the bruise or drinking more the following day, and also refraining from taking paracetamol.

My topic concerns the direct control that Tom (and only Tom) arguably has when he, say, reads an email and believes that the meeting starts at noon on that basis. We colloquially express this special form of agency in terms such as ‘making up your mind and keeping it made up’. It certainly doesn’t seem like we only influence our beliefs or indirectly control them as predicted consequences of our actions (pace Peels, 2017; Meylan, 2017, 2015). Why think this? Imagine that Tom appropriately investigates, discovers evidence that the meeting starts at 11am, but still believes that it begins at noon. Despite performing all the right physical actions, he still seems apt for criticism (perhaps even more than if he didn’t!). It’s not simply that Tom is derivatively responsible for his beliefs in virtue of his responsibility for his physical actions. The same also holds for mental actions. If you deliberate and judge that p without believing it then, as before, something has gone wrong rationally-speaking. It is not some kind of brute error in the same vein as faulty wiring.

(2) The claim isn’t that subjects can believe at will or for whatever reason they like. Whatever doxastic agency in fact amounts to, it isn’t going to be voluntaristic or intentional. There may well be pragmatic reasons for Tom to believe that the meeting starts at 11am not noon – if he believes the meeting is earlier than it is, then he’ll definitely arrive promptly (suppose Tom is habitually late). Yet, arguably, Tom can’t choose to believe that it will start at noon, and can’t believe on the basis of pragmatic reasons. At any rate, I don’t have this in mind when discussing doxastic agency.²

(3) I am interested here in the sort of agency which grounds epistemic responsibility. I can allow that there are other more minimal forms of agency and control; perhaps infants and penguins could be said to exercise something like this even though we wouldn’t criticise their beliefs. But in discussing doxastic agency, I have in mind specifically the sort of reflective control that grounds Tom’s responsibility for his belief.

So, I’ve introduced the claim on the table – that subjects have direct agency over their beliefs. But does this in fact hold, and if so, what makes it the case that subjects exercise this agency? These questions interrelate. If we can’t formulate a plausible and satisfying picture of what grounds doxastic agency, then we have all the more reason to deny that subjects possess it, or at least to the extent that we standardly assume they do.

2 Explaining doxastic agency: problems and puzzles

Here, I consider a first pass at explaining doxastic agency but argue that it still doesn’t suffice. I first introduce the account (Sect. 2.1) and a central remaining question (Sect. 2.2), before rejecting two potential solutions (Sect. 2.3).

2.1 A first pass

The most promising account of doxastic agency on the table appeals to dispositions to perform various mental actions which form and sustain the subject’s belief. After all, it’s natural to think that agency importantly relates to action.³ I’ll call this first pass the simple belief-sustaining account. The picture I have in mind grounds our doxastic agency specifically in dispositions to engage in reasoning – these especially seem most promising to secure direct control.⁴ So Sally possesses agency in believing that it will rain in being disposed to reassess the evidence that it will rain and potentially revise her belief in light of this.⁵ Perhaps Sally engages in a long period of deliberation, but it might also be that she just recalls a reason such as the weather forecast. Perhaps she never in fact engages in reasoning, but she could.

This looks especially appealing when compared to attempts to account for subjects’ agency in the way in which they form the belief, such as through deliberation and/or conscious judgement. (By ‘judgement’ I mean something like the mental event of affirming that p). I’ll call this the belief-formation account (e.g., Cassam, 2010; Peacocke, 1998; Shah and Velleman, 2005; Soteriou, 2005).⁶ For example, this approach locates Sally’s agency concerning her belief that it will rain in the way in which she initially forms it by considering reasons and concluding that it will rain. But just appealing to actual actions involved in belief-formation faces several worries. First, the belief-formation account excludes perceptual beliefs and also those the subject hasn’t explicitly deliberated about – e.g., we may well come to believe that it will rain upon checking the forecast without further thought. Nevertheless, as Chrisman (2018) writes, if subjects have agency over beliefs at all, they have it as much over these beliefs formed without explicit deliberation as those formed using it. But even if one didn’t explicitly deliberate concerning these unreflectively-formed beliefs, one could do so. Second, even when subjects have deliberated in the past, this isn’t to say anything about what sustains the belief now (Chrisman, 2018; Hieronymi, 2009a; Boyle, 2011). E.g., suppose Sally deliberated when forming the belief that it will rain but we now inform her of some counter-evidence. Here, Sally should change her mind and not simply reply: ‘don’t blame me, I formed the belief a while ago. I’m stuck with it now’. Even if past belief-formation methods are relevant, they are not the only thing that is. As Hieronymi writes, ‘my beliefs do not sit in my mind as last week’s lecture sits on my hard drive, recording what I once thought. My beliefs are rather my present, ongoing take on what is so’ Hieronymi 2009a, 176). A belief sustaining account, such as one appealing to the disposition to engage in reasoning and belief-revision, avoids these worries.⁷ Subjects possess the disposition even if they haven’t actually manifested it, and they can acquire, continue to possess, or lose this disposition at any point.

I ultimately have in mind a picture similar to McHugh (2017, 2013) who appeals to dispositions to engage in reasoning as part of a reasons responsiveness account of mental agency. Roughly, we have agency over our beliefs in virtue of being able to hold and revise them in light of reasons. We paradigmatically respond to reasons by reasoning, and are able to do so even if we happen to revise a particular belief unconsciously. The role of mental action and responding to reasons seems to be at least partly causal on this account. In fact, it’s worth flagging up a worry one might have with a causal account of this relation (the so-called ‘process’ account of doxastic agency).⁸ We might think that doxastic control could only be indirect on this account because of the causal intermediary, and that it would only amount to ‘getting oneself’ to have the belief. As such, forming a belief in this way would still be relevantly akin to influencing other people’s belief-formation. I will revisit this worry later. I would, though, also be open to related positions. One might prefer a picture akin to Hieronymi (2006, 2008, 2009a, b, 2011), according to which believing is an activity one engages in; reasoning in a sense constitutes believing. Or perhaps dispositions to engage in reasoning do not in themselves constitute agency, but ground the capacity to exercise agency which we manifest in causally sustaining the belief through reasoning. The key point I want to make is that mental agency importantly involves the disposition to engage in reasoning. While we might cash it out in different ways, a belief sustaining account provides a fundamental basis for doxastic agency.

In framing the topic in this way, I have had to leave other issues aside out of necessity and space.⁹ For example, I haven’t said anything about the role of voluntary choice in responsibility. But here I want to focus on the relation between our actions and agency given the worries we might have about how subjects can be active regarding a mental state (e.g., Setiya 2013). Also, in the practical domain, focussing on voluntary choice as the grounds of responsibility allows us to accommodate subjects’ responsibility for foreseen but unintended consequences, like if I spend too long in the restaurant even though I know my friend is waiting and will get bored. But this isn’t at issue here given that my aim is to understand direct doxastic control. I must also put aside the topic of moral luck. Perhaps one might think that lucky elements can bear on our moral evaluations. The possible examples differ, and my responses to these would depend on the precise cases at hand. But to the extent to which we deny that responsibility requires control to begin with, our overall project in understanding epistemic responsibility would look very different. Talking of control is already to assume a certain non-accidentality (and this is something I explore more later on).

2.2 Problems for the simple belief-sustaining account

Advantages aside, though, the simple belief-sustaining account (in any iteration) fails to fully cash out the way in which mental action grounds doxastic agency. Or at least, we need to say more to uncover the root of why it works and to provide a full and satisfying explanation.

In particular, we face the following puzzle which the account on its own doesn’t answer. Simply appealing to mental actions that bring about the belief isn’t to say how these amount to control. Suppose that doxastic agency (or the capacity for this agency) involves being disposed to perform a range of mental actions which form and sustain the belief that p.

In what way do these mental actions furnish the subject with control over her sustained token belief that p as opposed to simply happening to bring it about?

Following the account on the table, these actions play a causal or constitutive role in forming and sustaining our beliefs. But even if actions in some way bring about a certain state of affairs, this isn’t to say that the subject herself has exercised control over the state and can be credited with this. The following discussions focus on causation relations for the sake of simplicity, but we could also cash out similar examples in terms of constitution.¹⁰ One way of finessing this question further is to ask more precisely:

What makes it such that we don’t simply bring about the belief by accident in performing these mental actions?

For the belief to nevertheless be accidental would undermine the subject’s control.

In particular, we need to rule out cases in which the subject herself only brings about her beliefs by accident. It isn’t enough that there’s some non-accidental process behind her belief – after all, some of the following scenarios might involve machinations on the part of other people.

Subjects can bring about certain states by accident despite performing relevant actions. This is most obvious in the case of physical states. We can also find similar cases when subjects are trying to bring about a belief using indirect measures. Interestingly, though, it’s not clear there are parallel cases regarding deliberation and reasoning. I.e., if a subject reasons/ is disposed to reason about a given belief, and if this reasoning does indeed cause her to have that belief, then it seems that she will necessarily be responsible for it.¹¹ Indeed, this gives more cause to think that the control in question is direct. But even if reasoning does have this feature, the proponent of the belief-sustaining account needs to say why being disposed to engage in reasoning necessarily furnishes one with agency over the resulting belief.

There’s three aspects of our beliefs that we need the right kind of non-accidental relation to. How is it that being disposed to engage in reasoning furnishes subjects with control over belief, control over a sustained belief, and control over a belief that p? That is, how is it not just an accident that the actions give rise to a belief, a sustained belief, and a belief that p? The following examples perform a double shift. My goal at this stage is to illustrate the need to rule out accident-type cases, however we end up ruling them out. But they also help motivate my specific account and the appeal to self-awareness.¹²

2.2.1 Strand 1: Agency over *belief*

A subject’s actions might in some way alter state s without her exercising agency concerning s or being creditworthy for s. And this, it would seem, is because in performing these actions, the subject nevertheless brings about the state accidentally. (This is not to deny that there is some causal explanatory story that connects the actions and the state, but it does not involve the subject herself in the right way.) In the case of physical action, suppose Sally becomes strong as a result of carrying her shopping for miles, running, doing press ups etc. But perhaps Sally carries the shopping because she lacks a car, runs because she’s late for the bus and so on. Her strength is a happy accident, and it’s not clear we would praise her in this case. We’ll credit Sally with doing something, namely, performing various actions such as carrying her shopping and running. But we wouldn’t praise her for her resulting strength. Or at any rate, our ‘praise’ would be no more weighty than the way in which we ‘praise’ someone who happens to be tall. (To anticipate my later appeal to self-awareness, note that we especially wouldn’t credit Sally for this if she doesn’t connect these actions at all with becoming strong. Suppose she doesn’t even do so when prompted or in any sort of implicit way.)

Analogously, say that Sally wants to quit smoking and so visits a hypnotist. But the practitioner confuses Sally with a patient who instead wants to believe that open waters are safe, and accordingly gets Sally to believe this. As a result of visiting the hypnotist, Sally has acquired a belief; but, as with the above example, only accidentally and due to luck. Sally doesn’t appear to have exercised control over her belief qua belief. It’s not clear that we would, say, credit her in the same way we would credit someone who believes that water is safe having taken control of their phobia and approached the hypnotist for help.

In what way, then, does a disposition to engage in reasoning furnish one with agency over the resulting belief? After all, suppose Sally considers the reasons and then forms/maintains the belief that it will rain. The fact she now has a belief doesn’t seem to be a lucky accident, and we can accordingly credit Sally for her belief. But in what way does reasoning give Sally this specific control?

2.2.2 Strand 2: Agency over *one sustained token* belief

The first task, then, is to say how mental actions involved in reasoning relate to belief as a type; the second is to say how they furnish subjects with agency over sustaining one token belief. Recall that we are interested not just in subjects’ control in initially forming the belief, but also continuing to hold it. For example, say Sally believes that Smith is trustworthy. Following the account on the table, this belief is an exercise of Sally’s agency in that she is disposed to perform various mental actions. Suppose that Sally deliberates and comes to believe that Smith is trustworthy on Tuesday, reconsiders her reasons but reaffirms the proposition on Wednesday, and provides these and other reasons when asked by a friend on Thursday. Sally’s actions bring about and sustain one enduring belief – it is not that she stops holding the belief and then forms it anew each day.¹³ And indeed, the resulting belief is not merely accidental, and we can credit her for it.

Yet one can be responsible for originally acquiring a state, and indeed play a causal role in the fact that the state endures, without being creditworthy for the fact that it endures. For example, suppose Sally becomes strong in a non-accidental way having performed various actions. Still, she might be very inconsistent having acquired this strength. Say she exercises for a month, stops, starts again at the new year, etc. Suppose her strength stays stable in-between. Nevertheless, we wouldn’t praise her for being strong in a sustained way – this was just a happy accident. There’s a special kind of praise reserved for people who visit the gym every day and keep it up as opposed to yo-yoing.

We can also consider similar cases regarding attempts to exercise indirect control over one’s belief. Suppose that Sally visits a therapist because she wants to believe that open water is safe and receives weekly homework (e.g., dipping her toes in the water). The therapist warns her to practice regularly but Sally doesn’t. Maybe Sally happens to be highly susceptible even to this weak intervention, and does indeed acquire and sustain the belief. Still, we wouldn’t credit Sally herself with having sustained her belief – she was just lucky.

Why is it, then, that when Sally forms and sustains her belief through reasoning, we can credit her with this, i.e., with holding one continued belief? We need to do something to rule out cases in which her mental actions form and sustain one token belief by accident.

2.2.3 Strand 3: Agency over the belief *that p*

Third, in what way does reasoning furnish one with agency over a belief with a specific content? It is not just that we can credit subjects with holding a belief, but also, with holding the specific belief that p. Indeed, reasoning always seems to provide subjects with this specific control: when Sally forms and sustains the belief that it will rain through reasoning, we would praise/blame her for believing that it will rain. But why is this the case?

Although not an exact corollary, we might consider the following case involving physical action. Suppose that Sally tries to acquire strong biceps but performs the wrong sort of push up and instead develops strong forearms. As far as she is concerned, military-style press-ups (where one keeps the elbows particularly close to the body) are the only kind of press-up and require just a strong core and biceps. She has no idea at all that it exercises one’s forearms. We might credit Sally with a lot, and indeed praise her forearm strength insofar as it helps constitute her strength overall. But we wouldn’t credit her with strong forearms per se – this specific strength was accidental as far as she was concerned. As support, note how we would be hesitant to ask Sally’s advice if we wanted to improve at rock climbing (where forearm strength is key), and we wouldn’t praise certain aspects of her performance. Or to give another example, suppose that Sally follows a recipe for carrot soup. She indeed makes a soup, but gets confused between parsnips and carrots. We wouldn’t then credit her with making a parsnip soup, e.g., by praising her for the subtle interplay between the chilli and nuttiness of the parsnip.

And we can imagine similar scenarios concerning a subject’s efforts to control her own belief indirectly. Perhaps Sally visits a hypnotist because she wants to believe that open water is safe, but due to a mix up with patient records, instead ends up believing that smoking is bad. We would credit Sally with (indirect) control in coming to hold a belief, but not for coming to believe specifically that smoking is bad. Again, having a belief with this specific content was just accidental. And indeed, the content of the belief is important here. It isn’t simply that we credit people with having a true belief, as opposed to specifically the belief that p. In wanting to believe that water is safe, Sally wants to hold a true belief. Perhaps the hypnotist gave her the belief that smoking is bad in knowing this desire. Sally thereby has control over believing something true – but not the true belief that p. Yet this seems to be a departure from the norm.

So to sum up, our task for developing the simple belief-sustaining account is now to specify the way in which these mental actions furnish us with control and thus can serve as the grounds of epistemic responsibility, in the sense of ruling out merely accidental influence.

2.3 Answers rejected

Suppose, as I in fact think we should, that doxastic agency involves a disposition to engage in relevant reasoning. What is it about reasoning that furnishes subjects with agency over the resulting belief that p? As we have seen, it is not enough that reasoning causes or constitutes believing that p for a subject to have agency in this regard – this doesn’t rule out cases where she only accidentally brings about her beliefs.

I will now consider two possible answers, but neither seems to suffice (at least as they stand).

2.3.1 Option: intentions

Here we might consider the role of intention in physical action. This provides one way of distinguishing cases where Sally becomes strong as a result of her agency from ones where it is accidental. In the ‘good’ case, Sally becomes strong intentionally – because she decides to do so and visits the gym as a way of bringing this about.

It will help here to consider the way in which physical actions and activities can be complex and can be described in different ways. For example, suppose Sally turns on a light. Normally, she would successfully exercise agency in doing so and we would hold her accountable for turning on the light and not simply flicking a switch. E.g., we might complain: ‘why are you turning on the light when I’m trying to sleep – you’re so inconsiderate!’. And indeed, we respond legitimately even if Sally has only raised her arm at that point. But in what way are subjects accountable for performing a richly described action like this – turning on the light as opposed to flicking the switch? Because they act with a certain intention, and can provide that intention when asked. Relatedly, Hieronymi (2014) writes that controlling one’s physical actions seems to require being aware of the option of phiing, deciding that ought to phi, and deciding to phi; she terms these the ‘awareness’ and ‘discretion’ conditions on control.¹⁴

Perhaps, then, we might extend this model to belief. Maybe subjects engage in the mental actions involved in reasoning with the intention of thereby forming and sustaining the belief that p.

But, as discussed at the start, doxastic agency isn’t intentional. As Hieronymi (e.g., 2014) and McHugh (2013) observe, you don’t decide that you ought to p and intentionally bring it about. You can control your beliefs in this way, but that would be to exercise the indirect control I mentioned earlier, such as when Sally decides that she ought to believe that open water is safe and accordingly visits a hypnotist. In Hieronymi’s terms, this would only be to ‘manage’ your beliefs as you might a pot plant, or indeed a friend’s beliefs. The mental actions would only ever play a role in ‘tinkering’ with belief formation – setting yourself up to believe certain things and influencing your beliefs (Strawson, 2003, and extensively discussed in Vierkant, 2022). Also, we don’t set out to believe a certain content when we deliberate – at most we intend to form a belief, but not one in particular (e.g., Meylan, 2015, 2017).

Would it be enough that subjects intend to form a true belief? But we seem not to just have responsibility over having ‘what happens to be a true belief’ or ‘what is a true belief as far as we are concerned’, but our belief that p. Here I can say three things: (1) Sally could also intend to form a true belief in going to the hypnotist and requesting a true belief of their choosing. She might indeed acquire a true belief in this way, but this seems abnormal – and not just because she’s reliant on external help. (2) We can assess someone’s belief on the grounds of how the belief with that content coheres with their broader network of beliefs. (3) We would still hold subjects accountable and criticise them despite performing all the correct mental actions apt for producing a true belief (deliberation and the like) if they nevertheless judged that p was true and yet believed that not-p. And additionally, we seem to exercise doxastic agency on many occasions without engaging in stretches of intentional reasoning. Suppose I see the weather forecast when scrolling through the news and accordingly believe that it will rain. Here I may well take the forecast to be a good reason and judge that it will rain, but it isn’t that I intend to form a belief or engage in intentional reasoning.

2.3.2 Option: directed reasoning

Alternatively, we might appeal to reasoning as a directed activity. According to this thought, reasoning has an aim, and indeed one which importantly relates to belief. So, when performing the relevant mental actions, we aim at ‘forming a true belief’ (McHugh & Way, 2018) or at ‘settling the question whether p’ (e.g., Hieronymi, 2009a). Jenkins (2021b) explicitly grounds mental agency in ‘reasoning as an aim-directed activity’ where one relevant aim is that of ‘knowledge whether p’. Importantly, we might think that reasoning constitutively has this aim, thus explaining why successful deliberation always furnishes subjects with control over the resulting belief. But this picture still doesn’t suffice for at least two reasons (the second hints to my own answer).

(1) Suppose that reasoning does aim at forming and sustaining a true belief. Indeed, I think it does. Nevertheless, any given instance of reasoning aims at just that – forming a true belief. The subject doesn’t aim at forming one true belief in particular, i.e., the belief that p.¹⁵ But as we have seen, we can credit subjects with holding the belief that p, not just with holding a belief. It is not that subjects engage in reasoning and sustain a belief with that particular content by accident. We are not like Sally who visited the hypnotist because she wanted to believe that open water is safe and ended up with the belief that smoking is unhealthy by chance. We would credit Sally or having acquired a true belief, but just a true belief, and not the true belief that p; yet this case is unusual.¹⁶

(2) Additionally – to pre-empt my own account – at least some ways of construing the notion of ‘aim’ are too weak, and richer notions risk collapsing into the account I introduce later on. As such, the account fails to accommodate doxastic agency from the subjects’ own perspective. For example, McHugh and Way (2018) claim that in aiming towards holding fitting attitudes, subjects perform mental actions which accord with what will in fact bring about fitting attitudes, and are disposed to alter their reasoning rules if needed. For instance, subjects are disposed to use modus ponens when appropriate and only when appropriate. But simply appealing to dispositions to reason in accordance with this aim isn’t to say that subjects are aware of the aim or how what they are doing helps satisfy it. A subject could have such a disposition while having no idea that they are reasoning in a way that will bring about true beliefs. They might even think that they are not reasoning in this way and that they are not making up their mind about p.¹⁷ As such, we still haven’t provided an account of doxastic agency from the subject’s own perspective. Recall that agency, whether practical or epistemic, is something that the subject herself exercises; we hold her accountable for her beliefs.¹⁸

Following the above discussion, then, one might even have doubts about doxastic agency in the first place. The best account of doxastic agency on the table would locate it in dispositions to perform certain mental actions. Yet we need to do more to say how these actions furnish us with agency over a sustained belief that p as opposed to just happening to cause it. Given these difficulties, we might instead bite a bullet, even if we were initially sympathetic to the doxastic agency thesis. We might accept that subjects only have control over having a belief and not their belief that p. Or we might accept that, contra intuition, doxastic control is in fact intentional, and that subjects only have control over their beliefs in performing intentional mental actions which indirectly influence their beliefs.

3 The solution: agent’s awareness

We should nevertheless preserve the doxastic agency thesis and avoid biting any of the above bullets. Insights from the philosophy of action will help us. I agree with belief-sustaining accounts that exercising agency over believing that p in some way involves the disposition to perform mental actions involved in reasoning. But further, I propose that these mental actions furnish subjects with agency over the belief that p in part because the subject herself views what she is doing as exercising agency over her belief that p.

Let us start by recalling the complexity of physical actions and activities. In particular, I’m interested in the sorts of reflective actions that we often think of as intentional (perhaps there will be other actions to which this picture doesn’t apply). In what way can subjects be said to turn on a light and not just flick a switch? We can answer: because the subject sees herself as turning on the light. Here we can draw on Anscombe’s (1957) discussion of the question ‘why?’ regarding intentional action.

When someone is, say, turning on the light, she is able to say that she is turning on the light and she knows that she is doing this. For Anscombe, the subject is also normally able to say why she is acting. But the action ceases to count as intentional under that description if the question “why?” ‘is refused application’, i.e., if the subject doesn’t even see the precise question as one that she can or should answer (1957: 11). And one way of circumventing the question in this way is to reply that you didn’t know that you were turning on the lights at all. This isn’t to say that the subject isn’t nevertheless doing something intentionally, but it would be under a different mode of description – perhaps just flicking switches. The point is that doing something under one mode of description rather than another (e.g., turning on the light as opposed to merely flicking switches) seems to require taking yourself to be engaging in that activity so described. This observation starts to help us with the question of how it is that our actions can furnish us with control over a state of affairs, and in particular, how to distinguish these cases from those of merely accidental influence. If Sally replies that in flicking switches she is turning on the light, we would be happy to say that she is indeed turning on the light as opposed to just flicking a switch (or at least, we would if she is successful). As such, it isn’t a mere accident that Sally ends up illuminating the room; we would credit her with this and thank her if we were so inclined.¹⁹

What broad message can we take away from this? After all, Anscombe’s point is very much embedded in her metaphysics of intentional action – and I precisely don’t want to appeal to intentions in understanding doxastic agency. Abstracting away the role played by intentions in the cases traditionally discussed, I think more broadly, phi-ing as opposed to psi-ing requires being at least disposed to become aware of phi-ing as opposed to psi-ing. In the case of intentional action, this would be roughly because acting under one description is to act with the relevant intention, and we are thereby aware of acting with that intention. But more generally, acting under one description as opposed to another is for that action to be subject to certain success conditions instead of others. On one interpretation of Anscombe, ‘intention sets a standard of success for what does’ (Setiya, 2022). So in turning on a light as opposed to merely flicking switches, and in taking ourselves to be turning on a light switch as opposed to flicking switches, we recognise that we are subject to a particular standard of success – namely, that the light turns on.

Suppose we accept this broad moral – that performing an action requires seeing yourself as performing that action under the relevant mode of description. This observation also helps us when accounting for doxastic agency. In what way do mental actions furnish subjects with agency over the beliefs that these actions form and sustain? I propose the following self-awareness belief-sustaining account of doxastic agency.²⁰ This is a constitutive account, about what it is to exercise doxastic agency:

Subjects exercise agency over their belief that p in being disposed to perform a range of mental actions in reasoning which form and/or sustain the belief, where they are aware of these actions as part of reasoning and exercising agency over their belief that p.²¹

Seeing yourself as switching on a light can make the difference between counting as intentionally switching on the light and just flicking a switch. Similarly, I propose that seeing yourself as exercising agency over your beliefs helps make the difference between exercising agency over your belief and just engaging in various mental actions that happen to bring it about. It is worth emphasising that I take this self-awareness to be part of what it is to engage in reasoning (where I understand reasoning to be a personal level activity that the subject herself performs). It makes it such that the mental actions furnish one with control over the belief. I therefore doubt that it would be enough to just appeal to the capacity to become aware of one’s mental agency in a purely contingent manner, e.g., through inferring it from one’s behaviour and others’ testimony. It would have to be grounded in a tight and necessary way and also be distinctly first-personal. I’ll say more later on. Also, by ‘awareness’ at this point, I have in mind something like the Anscombian picture: the subject is in a position to form a belief that she is exercising this agency, which then she can express to a listener.

To give an example, suppose that Sally believes that it will rain. She originally formed the belief by considering whether it will rain and, on the basis of facts such as the grey clouds, concluded that it will indeed rain. Later, Tom asks her if it will rain; Sally recalls the grey sky and so replies that it will. Tom, though, looks sceptical and says that he read in the newspaper this morning that it will be sunny. Sally considers her reasons again; she only saw the clouds moments ago so she thinks this constitutes a stronger reason. As such, she continues to believe that it will rain. The account of doxastic agency on the table says the following. Sally exercises agency in believing that it will rain. She does so in virtue of performing mental actions involved in reasoning (initially deliberating, and reconsidering the reasons when later prompted by Tom) and being aware that these actions are related – namely, they are all part of reasoning and exercising control over her belief that it will rain. It is not that Sally reconsiders her reasons blindly and has no idea whatsoever about why her belief endures. More precisely, she is not nonplussed about why she believes that it will rain as opposed to, say, desiring it. Or why she believes specifically that it will rain as opposed to believing that it won’t rain, or indeed, some unrelated content. Or why her belief endures all this time and will continue to do so. Rather, Sally grasps the way in which her mental actions relate to her belief that it will rain and is accordingly responsible for it.

I should emphasise that under my account, being in a position to ascribe agency is necessary but not sufficient for possessing it. For example, we may well also require that the mental actions play a causal role in sustaining the belief. I can thus allow that subjects might think that they are exercising agency over a belief without actually doing so, such as in the case of confabulation.²² In fact, my thought is that the self-awareness makes it such that the relevant mental actions can furnish us with agency over our beliefs. So it is not that the mere self-ascription of doxastic agency furnishes us with agency in and of itself. Rather, it renders the mental actions as being of the right kind to form and sustain our beliefs, and to preclude cases of accidental influence.

Before continuing, I’ll say something about how this proposal fits with the general moral I drew earlier from the ‘Anscombean’ picture. I suggested that, abstracting from the specifics of intentional action, seeing yourself as performing an action under a certain description helps make it the case that you perform that specific action because it is to recognise certain standards for success. Exercising agency concerning a given belief is not to intend to have that belief, or even to aim to have it. But we can still identify success conditions elsewhere. In responding to reasons, your belief ought to match those reasons appropriately. If all the evidence indicates that p is true, and especially if you judge that p, then you ought to form the belief that p or continue to hold that belief. In seeing yourself as making up your mind and keeping it made up with regards to your belief that p, you recognise a certain success condition on those mental actions. Even if you don’t initially see making up your mind in terms of success conditions, the thought is that you will be able to do so if you reflect on what this amounts to. And plausibly, recognising that your action is subject to this standard is a way of making it the case that you are in fact subject to it – i.e., that you are performing an action under one description as opposed to another. This is the case even though we do not set this standard ourself by intending to form the belief that p in judging that p.²³

I will now say why this account is independently plausible before revisiting the starting puzzle. I further develop and explain the picture in response to possible objections in § 4.

3.1 Independent plausibility

Let me note several points.

• a. By ‘agency over our beliefs’ I am referring to what is colloquially expressed as ‘making up our mind and keeping it made up’. When I say that subjects will be in a position to form relevant self-ascriptions, that isn’t to say that they would form beliefs using the term ‘doxastic agency’. Rather, it would be for the subject to say something like: ‘I am currently making up my mind about p’ or that ‘I am thinking again about my belief that p’. So, returning to the above example, Sally would be in a position to believe that her reasoning is part of making up her mind about the rain, and, later, part of reconsidering her belief that it will rain. What ‘making up your mind and keeping it made up’ amounts to more precisely will depend on one’s prior commitments, such as the role one gives to either causal or constitutive relations.

• b. We can contrast cases in which subjects seemingly exercise doxastic agency with ones where they don’t, and note the role that self-awareness plays in this difference. First, compare reasoning about your belief that it will rain with simply rehearsing a friend’s train of thought when imagining why they believe the same thing. This would involve the same inner speech, and yet we nevertheless expect subjects to recognise when they are simply rehearsing ideas. And second, compare the difference between our everyday beliefs and Bonjour’s (1980) clairvoyant Norman. Norman believes that the president is in town as a result of a reliable clairvoyant ability that he is ignorant of. As such, the belief just happens out of nowhere as far as he is concerned. But our everyday beliefs don’t come as a surprise to us and do not simply happen upon us. Rather, our beliefs seem explicable to us: we grasp that we have these beliefs specifically because we form and continue to hold them.²⁴ This isn’t to say that Norman can’t come to have a normal agential relation to his belief. Perhaps as time goes on, he will respond to evidence bearing on this belief. Norman will consider reasons if prompted and either continue to hold or drop his belief. But in this case, we would expect his belief to become explicable to him as he takes on this new agential relation.

• c. Note the way in which subjects reason about several things at once. They grasp what they are doing, and which trains of thought relate to which beliefs. And indeed, this is the case whether they are deliberating for the first time about their beliefs or reconsidering existing ones. At any one time, Sally might be thinking about the best combination of ingredients for her sourdough bread and reconsidering her belief that Smith is a good politician. For example:

What would work with roasted red pepper? Sun dried tomatoes are too similar. Smith hasn’t dealt with the current crisis very well. But the economy used to be very strong under her. Pumpkin seeds? That would be a good contrast. Yes, pepper and pumpkin seed would be delicious. Still, it’s easy to rule well in good conditions. Smith isn’t a good politician after all.

Yet Sally doesn’t get confused between the different sets of premises. These are two distinct trains of thought, and in switching between the two, Sally nevertheless seems to grasp what she is up to with each one.²⁵

3.2 Revisiting the puzzle

Let’s revisit the puzzle facing simple belief-sustaining accounts of doxastic agency. Namely, if we appeal to a range of mental actions, we need to say how they relate to the belief that p in order to furnish the subject with (direct) control over it, such that the resulting belief is not simply a happy accident. Let us recall my initial question’s three strands. (1) How do the mental actions relate to belief as a type? (2) How do the mental actions relate to a single token belief? The third question is especially important: (3) How do the mental actions relate to a belief with a particular content? Recall the other options for explaining the way in which mental actions furnish subjects with agency. We might say that reasoning is a directed activity which aims at forming true beliefs. Yet this doesn’t account for the subject’s agency concerning a belief with a specific content.

I’ve argued that subjects perform these mental actions as part of reasoning, where they are aware of themselves as exercising agency over their beliefs. This awareness plays a constitutive role in individuating the subject’s activity as being of the right kind. So even if we want to say that the mental actions play a causal role in forming and sustaining one’s beliefs, the subject must see what she is doing as exercising doxastic agency for the causal relation to count as one of the right kind. As a result, the subject performs actions that are specified such that by their very nature they are apt to furnish one with control over the relevant belief as opposed to just resulting in it accidentally. Subjects are aware of themselves as exercising agency over one token belief that p. This is to be aware of exercising agency specifically over a belief,²⁶one token belief, and the belief that p. Indeed, they are actions that the subject herself performs under this description (even though the actions are not fully intentional under this description). As such, the belief is not a mere accident.

This will be fairly straightforward in the case of sustaining a previously formed belief. The subject will be aware of having a belief with a specific content, and can therefore be aware of exercising agency over her sustained belief that p. But what about when the subject is initially deliberating and hasn’t yet formed the belief? Still, I think that as subjects deliberate and certain options become more appealing, subjects become gradually aware of the specific belief they will end up forming. Even if they haven’t yet fully made up their mind, they can still say what position they are ‘leaning towards’.

Recall also that I’m concerned here with direct doxastic agency. The thought is that self-awareness individuates the mental actions such that they constitute those of making up one’s mind and keeping it made up without the subject having to do anything else. This is the case even if we think that the mental actions cause and causally sustain the belief. This can still be something that we ourselves do directly even if there is a causal intermediary. After all, we can directly turn on the light in virtue of flicking a switch even though the electronics rely on many causal processes. Compare this to turning on the light indirectly, such as by asking a friend to do so.²⁷

We can also now note the way in which the ‘aim’ view could account for subjects’ responsibility for forming and sustaining the belief that p, but only by falling into mine. Its proponents might have a rich sense in mind when they say that reasoning aims at, say, settling the question whether p. They might think that subjects are aware of the aim and of what they are currently doing in trying to fulfil it. But under this interpretation, the ‘aim’ view sufficiently resembles my own that I accept it.

4 Objections and replies

I will now consider several objections to my account.

4.1 Overintellectualisation?

We might worry that my account overintellectualises rational agency. While adult humans can engage in explicit reasoning and reflect on this process, they rarely do – perhaps this contradicts my claim that doxastic agency is importantly self-conscious.

4.1.1 Reply

Here I will say several things. First, the account doesn’t require that subjects possess very complicated concepts. Doxastic agency as I understand it here amounts to the everyday notion of making up one’s mind. Second, I’m not saying that subjects always engage in reasoning about their beliefs and are thus always aware of exercising agency. After all, we might have a great many beliefs that rarely, if at all, become occurrent (like the belief that Jupiter is the largest planet). And I do not require that for each belief that p, subjects also believe that they have agency over this belief. The claim is that subjects are disposed to perform certain mental actions and that subjects are in a position to form the relevant meta-belief. Third, while I have argued that subjects see themselves as exercising doxastic agency in performing mental actions, this is not in itself to say that they must endorse this process. That is, my account differs from Frankfurt’s (1971) ‘higher order endorsement’ picture of agency and free will. I happen to think that agency also in fact does require higher-order endorsement, but one can accept the paper’s main point without this further commitment.

4.2 Grounding the self-ascription

I have suggested that subjects exercise doxastic agency over their beliefs in performing a range of mental actions that form and sustain the belief, and in being aware of these actions as part of exercising agency over it. So far, I have cashed out ‘awareness’ in terms of being in a position to form meta-beliefs self-ascribing this agency. But what grounds these meta-beliefs, i.e., in what way are subjects in a position to form them? My objector might question whether I can satisfactorily account for this. After all, I can’t appeal to an intention to form and sustain one’s belief that p which one then reflects on. (Anscombe’s own account regarding action rests on the nature of intention.) It also seems doubtful that subjects learn of their doxastic agency by inferring it, or by using an inwardly-directed detection mechanism to observe their mental actions akin to the way in which their ‘outer eye’ perceives objects in the world. As has been oft observed, we consider p itself when we are deliberating, not our mental states.²⁸

4.2.1 Reply

I think that subjects learn of their agency in virtue of exercising it, and certainly it seems plausible that the self-ascription in question would have to be necessarily related to the mental actions in this sort of way. There’s different options and I’ll say something about my preferred one to start with. It’s also worth noting that the question of how I know that I am exercising doxastic agency regarding a belief is distinct from the question of how I know that I have that belief. Perhaps one might think that our knowledge of the belief is not in itself grounded agentially, but is perhaps, say, acquired through introspection.

The grounds of practical knowledge and how to make sense of the thought that it consists in ‘intention in action’, is itself debated. In both the practical and epistemic case I want to appeal to the role of experience, although I encourage readers to consider how their own preferred account would extend. Let us return to the philosophy of action and the notion of ‘agent’s awareness’. We might have one of two things in mind when saying that a subject is aware of something. For example, she could perceive a cake and experience it being there, or she could have learnt about it from a friend. That is, we can talk about awareness in either an experiential or epistemic sense. Indeed, awareness in the experiential sense can ground our epistemic access – Sally can come to believe that there is a cake because she perceives it. So far I have discussed an awareness of acting in an epistemic sense – the subject is in a position to say what she is doing and form beliefs about it. But we can also pinpoint an experience concerning one’s actions, and a sense of agency.²⁹ So the thought goes, when a subject raises her arm, or even performs mental actions like calculating and judging, she has an experience concerning what she is doing and her agency in doing so. Indeed, it seems that she has this awareness even if she has her eyes closed and cannot see what she is doing. It is not that she must observe her hand’s movement; rather, she experiences her action in being the one performing it. Similarly, I think that subjects can form beliefs about their doxastic agency in virtue of experiencing doing so, in being engaged in forming and sustaining their beliefs. This process occupies their attention without being its object (see Peacocke 1998).

Here, I can emphasise extant literature that already argues that we can enjoy a sense of agency when engaging in certain mental actions involved in reasoning. E.g., Peacocke lists the following as examples of mental events towards which we seem to have a sense of being active: ‘decidings, judgings, acceptings, attendings to something or other, calculatings, reasonings, tryings’ (2007: 361). See also O’Brien’s (2007) discussion of agent’s awareness of judgement, Campbell’s (1999) appeal to a sense of agency in reasoning which he uses to explain thought insertion cases, and Doyle’s account of the sense of agency in thinking (2019, 2022).³⁰ I am simply trying to extend this by saying that we can be aware of the relevant conscious mental actions in a particular way – as of exercising agency over our beliefs. I.e., subjects experience these actions as part of making up their mind.

Alternatively, one might look to other options away from the philosophy of action literature, and just appeal to awareness in a purely epistemic sense – being in a position to form relevant metabeliefs – and provide a different story of how the beliefs would be grounded. One option would be to appeal to transcendental entitlements along the lines of those in Moran (2003, 2001). Perhaps we are warranted in believing that we are exercising agency regarding our belief that p because of the very fact that being disposed to believe this is a precondition on exercising doxastic agency. Or perhaps the metabeliefs aren’t warranted at all per se, but count as knowledgeable in virtue of constituting what they ascribe.³¹ I, though, think that these sorts of self-ascriptions must be warranted in a way that’s understandable from the subject’s perspective.

5 Conclusion

This paper has helped defend the doxastic agency thesis by providing an account of the way in it would be possible, and in particular, an account by which mental action furnishes subjects with agency over their belief that p. Subjects exercise doxastic agency in being disposed to perform a range of mental actions that form and sustain their belief that p and in being aware of these actions as part of reasoning and exercising agency over their belief that p. Appealing to self-awareness in this way helps us to explain how subjects can be held accountable for a sustained belief that p and not simply for performing individual mental actions.

Having answered one question, though, others remain. In discussing the grounds for forming the relevant meta-beliefs, I appealed to a sense of agency regarding our beliefs. But what psychological structures underpin this? As mentioned earlier, philosophers and psychologists already discuss subjects’ sense of agency concerning mental and physical action, and practical deliberation. And indeed, many philosophers have studied the way in which subjects come to know their beliefs. But a sense of agency concerning beliefs – which are states as opposed to actions – remains relatively unexplored. This is a particularly exciting task since many discussions of agent’s awareness are specific to the case of intentional action (e.g., efference copy/comparator model accounts). Indeed, agent’s awareness is often seen as ‘intention in action’.³² Formulating an account of how this extends will help to further cash out my basic proposal, and also develop the best account of the sense of agency simpliciter.