Attitudes and action: against de se exceptionalism

ABSTRACT

De se exceptionalism is the view that de se attitudes pose a distinctive problem for traditional theories of propositional attitudes. A recent argument for de se exceptionalism attempts to prove that the distinctive problem of de se attitudes has something to do with the role of de se attitudes in explaining actions. The argument is based on a case where two subjects seem to believe and desire all the same propositions but perform different actions. This is the most promising argument for de se exceptionalism in the current literature. This paper rejects the argument by proposing a novel model of action explanation, according to which agents perform what they consider to be the best actions. I will argue that the case, based on which the recent argument for de se exceptionalism is developed, is philosophically interesting. Traditional theories of propositional attitudes can accommodate the case easily. I will conclude that de se exceptionalism should be rejected.

KEYWORDS:

• First person
• action explanation
• propositional attitude

1. Introduction

De se exceptionalism is the view that de se attitudes pose a distinctive problem for the traditional theories of propositional attitudes. A recent argument for de se exceptionalism attempts to prove that the distinctive problem of de se attitudes is related to their role in explaining actions. This paper rejects this argument by proposing a novel model of action explanation.

Classical cases that motivate de se exceptionalism usually involve a subject who suffers de se ignorance – i.e. who has information about herself but does not know that it is about herself.¹ These de se ignorance cases are often thought to pose a problem for the traditional view of propositional attitudes. An increasingly popular view claims that the arguments based on these cases are not persuasive. This view is called ‘de se scepticism’.² According to de se scepticism, the problem posed by de se ignorance cases has nothing particularly to do with de se attitudes; it is merely an instance of a more general problem facing the traditional view of propositional attitudes (e.g. Frege's puzzle).

In response to the sceptical challenge, many recent exceptionalists argue that we should focus on cases of a different type. According to them, the special problem of de se attitudes is related to cases where two subjects seem to have all the same relevant attitudes (according to the traditional view of propositional attitudes) but perform different actions.³ For example, a case that many recent exceptionalists consider is the bear attack case, first discussed by Perry (1977):

Bear attack.

Ann and Bill are walking in the woods when a bear starts chasing Ann. Ann and Bill both realise that the bear is about to attack Ann, and they both want the bear to leave. They act differently. Ann curls up into a ball and plays dead. Bill, who witnesses the attack from a distance, attempts to distract the bear by shouting at it. Perry (cf. 1977, 494)

These exceptionalists argue that the traditional view of propositional attitudes implies that Ann and Bill share all the relevant attitudes – some of them are Ann's de se attitudes, but this is in conflict with the fact that Ann and Bill perform different actions. These exceptionalists claim that we should reject (and only reject) what the traditional view claims about de se attitudes to avoid the problem in question.⁴

This type of argument is the most promising argument for de se exceptionalism in the current literature. In this paper, I am going to reject this argument and maintain the traditional view of propositional attitudes. I will consider two alternative explanations of the bear attack case. According to the first explanation, Bill does not perform the relevant action because the relevant action is not open to him. According to the second explanation, Bill does not do what Ann does only because he thinks there is a better action. The two alternative explanations are both compatible with the traditional view of propositional attitudes.

The first proposal is discussed by Cappelen and Dever (2013), which I find wanting. My own proposal will be the second one. I will first develop an intuitive model for action explanation, according to which agents perform what they consider to be the best actions. The model will be used to account for the bear attack case. If I am right, then, unlike de se ignorance cases (which pose a problem that can be replicated without any de se attitudes involved), the bear attack case and other similar cases create no problems at all; they are philosophically uninteresting. All that these cases illustrate is the following platitude: agents will perform what they consider to be the best actions.

Here is the plan. Section 2 gives some preliminaries. Section 3 explains the alleged problem posed by the bear attack case. Section 4 presents the exceptionalist solution. Section 5 discusses and critiques Cappelen and Dever's strategy. In Section 6, I will develop my own strategy. In Section 7, I will compare the three strategies I discuss in this paper. I will conclude that my strategy should be preferable, and that the argument for de se exceptionalism based on the bear attack is not persuasive.

2. Some preliminaries

First, by ‘de se attitudes’, I mean attitudes that their holders could usually express or self-report by using first-person pronouns, such as ‘I’ and ‘me’.⁵ For example, the belief that Ann could expressed by saying, ‘The bear is about to attack me’, is a de se belief of Ann's.

Second, the traditional view of propositional attitudes, which is the target of the exceptionalist argument discussed in this paper, contains four claims. The first three claims are as follows:

Binarity.	Propositional attitudes are two-place relations between subjects and propositions.
Shareability.	Propositions are shareable entities. That is, it is possible for different subjects (at different times) to bear the same type of propositional attitude to the same proposition.
Absolutism.	Propositions are absolute in their truth values. That is, whether a proposition is true or false depends only on the objective state of the world.

According to the traditional view, propositional attitudes are relations between subjects, on the one hand, and shareable, absolute propositions, on the other. This view is neutral on the nature of propositions. It is consistent with the view that propositions are Fregean or Russellian structural entities or sets of possible worlds.

The fourth claim that the traditional view endorses concerns the role that propositional attitudes play in action explanations. We often try to explain one's actions by attributing to one some propositional attitudes. For example, when it comes to the question, why did Cara cook for Dia, one might give the following answer: Cara cooked for Dia because she wanted Dia to be happy, and she believed that if handmade dishes are provided to Dia, then Dia will be happy. The explanation can be taken to consist of two facts: the fact that Cara holds the desire that Dia be happy, and the fact that Cara has the belief that if Dia eats handmade dishes, then Dia will be happy.

Psychological action explanations require the existence of law-like connections between the relevant attitudes and actions.⁶ Cara's action is explained by the fact that Cara wants Dia to be happy and believes that if handmade dishes are provided to Dia, then Dia will be happy. This is true only if there is a regular link that connects Cara's attitudes in question with the type of action that Cara performs. This means that if anyone x has the same attitudes as Cara, then, other things being equal, x will do what Cara does.

By contrast, suppose that Cara also wants to annoy Evan, and she believes that if Evan is pranked, he will be annoyed. There are no regular connections between this pair of Cara's attitudes and the action that one cooks for Dia. In other words, it is possible that an agent also wants to annoy Evan and also believes that if Evan is pranked, he will be annoyed, but the agent does not cook for Dia (even if all other things are equal between this agent and Cara). Since there is no regular connection between this pair of attitudes and the action that one cooks for Dia, the fact that Care holds them does not explain the fact that she cooks for Dia.

Note that the regulation between Cara's attitudes about Dia and the action that Cara performs is a ceteris paribus law, which involves an ‘other things being equal’ clause. It is difficult to give a complete description of this clause. However, we can at least say that in this particular case, other things are not equal between some agent and Cara if the agent, unlike Cara, does not know how to cook or is reluctant to cook.

To generalise this point, we have the fourth claim that the traditional view accepts:

Explanation.

Suppose that x holds attitudes $A_0,A_1,\dots ,A_n$ explains the fact that x performs an action token under the action type ϕ. Then necessarily, for any y, if y holds $A_0,A_1,\dots ,A_n$, then, other things being equal, y will perform an action token under type ϕ.

Now, with all the relevant claims presented, we are in a position to explain the alleged problem that some recent exceptionalists think is posed by the bear attack case.

3. The alleged problem of de se attitudes

Consider the bear attack case again. In this case, Ann and Bill both realise that the bear is about to attack Ann, but they act differently: Ann curls up into a ball, and Bill shouts from a distance.

Let's focus on Ann's attitudes and action for the moment. Presumably, the fact that Ann curls up into a ball is explained by the fact that Ann holds the attitudes $B_0$, $B_1$, and D, where $B_0$ is Ann's de se attitudes:

• $B_0$: the belief that Ann could express by saying, ‘The bear is about to attack me’;

• $B_1$: the belief that Ann could express by saying, ‘If I curl up into a ball when the bear is about to attack me, then the bear will leave’;⁷

• D: the desire that Ann could report by saying, ‘I want the bear to leave’.

Now, suppose that the traditional view of propositional attitudes, which commits to Binarity, Shareability, Absolutism, and Explanation, is true. Then, Bill shares all three attitudes. Consider the attitude $B_0$ first. Let's assume that $p_0$ is the propositional content of $B_0$. Then, according to Binarity, $B_0$ is a two-place relation between subjects and $p_0$. This means that anyone can hold $B_0$ as long as she bears the relation of belief to $p_0$. But, by Shareability, it is possible that Bill bears the relation of belief to $p_0$. Therefore, it is possible that Bill holds $B_0$.

What would Bill believe if Bill holds $B_0$? Given Absolutism, the following answer seems correct: if Bill holds $B_0$, Bill would hold a belief that Bill could express by saying, ‘The bear is about to attack A’, where ‘A’ is a linguistic expression that refers to Ann. Here is why: Given Absolutism, $p_0$ (the propositional content of $B_0$) has an absolute truth condition. But $p_0$ is the proposition that Ann could express by saying, ‘The bear is about to attack me’. Therefore, the absolute truth condition of $p_0$ presumably is this: $p_0$ is true relative to a world if and only if the bear is about to attack Ann in the world. Therefore, for Bill to believe $p_0$ is for Bill to believe a proposition that has the same truth condition. This means that there is a linguistic expression ‘A’ such that ‘A’ refers to Ann (in the right context), and for Bill to believe $p_0$ is for him to hold a belief that he could express by saying, ‘The bear is about to attack A’. For our purposes, let's just assume that ‘A’ is the indexical expression ‘you’.⁸

Now, it is reasonable to ascribe the attitude $B_0$ to Bill in the bear attack case. He seems to hold the attitude when he realises that the bear is about to attack Ann. The same goes for the other two attitudes that Ann has: it reasonable to ascribe to Bill in the bear attack case the belief $B_1$ (i.e. the belief that Bill could express by saying, ‘If you [addressing Ann] curl up into a ball when the bear is about to attack you, then the bear will leave’) and the desire D (i.e. the desire that Bill could report by saying, ‘I want the bear to leave’). But, then, if we accept the following claim, the bear attack case seems to pose a problem for the traditional view of propositional attitudes:

(i)	$B_0$, $B_1$, and D are all the relevant attitudes that explain Ann's action.

The problem is this: the traditional view of propositional attitudes implies that:

(ii)	Bill also holds $B_0$, $B_1$ and D.

From (i) and (ii), we can infer that:

(iii)

Bill shares all the relevant attitudes that explain Ann's action.

Which action is the one that is explained by Ann's holding $B_0$, $B_1$ and D? Well, it is tempting to claim that:

(iv)	The action that is explained by Ann's holding $B_0$, $B_1$ and D is the action, ‘curling up into a ball’.

Then, by (iii), (iv) and Explanation, we can infer that:

(v)	If other things being equal, Bill will also curl up into a ball.

But it seems that:

(vi)	There are no relevant differences between Ann and Bill, due to which, other things are not equal between them.

For example, both Ann and Bill can curl up into a ball. Bill does not struggle, mentally or physically, to perform this action.

But, then, from (v) and (vi), we can infer that Bill will also curl up into a ball. However, this contradicts the following fact:

(vii)

In the bear attack case, Bill does not curl up into a ball.

The traditional view of propositional attitudes seems to run into trouble.⁹

Note that we make three assumptions here:

(i)	$B_0$, $B_1$ and D are all the relevant attitudes that explain Ann's action.
(iv)	The action that is explained by Ann's holding $B_0$, $B_1$ and D is the action, ‘curling up into a ball’.
(vi)	There are no relevant differences between Ann and Bill, due to which, other things are not equal between them.

The inferences of the argument are all valid. Therefore, if we accept assumptions (i), (iv) and (vi), then we should conclude that there is a problem of de se attitudes, and the traditional view of propositional attitudes is the culprit.

4. The exceptionalist solution

Exceptionalists accept all three assumptions of the argument presented in the last section, and they conclude that we should revise the traditional view of propositional attitudes to avoid the alleged problem of de se attitudes. Furthermore, they believe that the problem is specific to de se attitudes, and, therefore, to avoid the problem, we should only revise what the traditional view says about de se attitudes.

There are many revised theories that meet these criteria. In this section, I will mainly focus on one of the most influential strategies: the centred-worlds account.¹⁰

Remember that the traditional view consists of Binarity, Shareability, Absolutism and Explanation. The centred-worlds account accepts Binarity, Shareability and Explanation, but it rejects Absolutism. The theory replaces Absolutism with the following thesis:

De se Relativism.

While the contents of non-de se attitudes are absolute, the contents of de se attitudes are relative: they could be true for one individual but false for another.

In particular, the centred-worlds account builds on the possible-worlds account, but substitutes possible worlds with centred (possible) worlds, which can be represented by a pair $⟨w,x⟩$ where w is a possible world and x is the subject at the centre of w.

Consider, for example, Ann's de se belief that she could express by saying, ‘The bear is about to attack me’. On the possible-worlds account, the content of Ann's de se belief is a possible-worlds proposition. If it is true (or false) relative to an individual in a world, it is true (or false) relative to any individual in the same world.

However, according to the centred-worlds account, the content of Ann's de se belief is the centred-worlds proposition {$⟨w,x⟩$: the bear is about to attack x in w}. The proposition does not have an absolute truth value. It might be true relative to an individual in a world but false relative to a different individual in the same world. For example, let $w_1$ be the world described in the bear attack case. Then the centred-worlds proposition {$⟨w,x⟩$: the bear is about to attack x in w} is true relative to $⟨w_1,$ Ann〉 but false relative to $⟨w_1,$ Bill〉.

On the possible-worlds account, a subject x believes a (possible-worlds) proposition in a possible world w if that proposition is true relative to every possible world that is doxastically accessible to x from w. However, on the centred-worlds account, x believes a (centred-worlds) proposition in w if that proposition is true relative to every centred world $⟨v,y⟩$ that is doxastically accessible to x from w. To say that $⟨v,y⟩$ is doxastically accessible to x from w is to say that it is compatible with what x believes in w that v obtains and y is identical to x.

Advocates of the centred-worlds account often wish to have a uniform account of propositional attitudes. They do so by maintaining that all propositional attitudes have centred-worlds contents. But while the contents of one's non-de se attitudes have boring centred-worlds contents, the contents of one's de se attitudes are interesting centred-worlds propositions. A boring centred-worlds proposition p is a centred-worlds proposition where the centres are irrelevant – i.e. for any w, x, and y, $⟨w,x⟩\in p$, if and only if $⟨w,y⟩\in p$. A centred-worlds proposition is interesting if it is not boring – i.e. if there are some w, x, and y such that $⟨w,x⟩\in p$, but $⟨w,y⟩\notin p$.¹¹ For example, the content of Ann's aforementioned de se belief is the interesting centred-worlds proposition {$⟨w,x⟩$: the bear is about to attack x in w}; the content of Ann's non-de se belief that the bear is about to attack someone is the boring centred-worlds proposition {$⟨w,x⟩$: the bear is about to attack someone in w}, which is equivalent to the possible-worlds proposition {w: the bear is about to attack someone in w}.

The centred-worlds account avoids the alleged problem of de se attitudes posed by the bear attack case. This is because the theory implies Bill does not share all the relevant attitudes that explain Ann's action. In particular, the theory implies that Bill does not hold $B_0$ (i.e. the belief that Ann could express by saying, ‘The bear is about to attack me’).

To see this point, consider: according to the centred-worlds account, $B_0$, which is A's de se belief, is a two-place relation between subjects and the centred-worlds proposition {$⟨w,x⟩$: the bear is about to attack x in w}. Therefore, for Bill to hold $B_0$ is for him to bear the relation of belief to the centred-worlds proposition, but Bill bears the relation of belief to the centred-worlds proposition if and only if Bill holds a de se belief that he could express by saying, ‘The bear is about to attack me’. However, in the bear attack case, Bill does not hold this de se belief.

The centred-worlds account is not the only de se account of de se attitudes that can accommodate the bear attack case. There are two other notable accounts of de se attitudes that have this feature – the Fregean account and the Perrian account.¹²

First, according to the Fregean account, de se attitudes are two-place attitudes between subjects and propositions that are absolute but not shareable. When Ann believes that she herself is about to be attacked by the bear, she believes a proposition that cannot be shared by Bill.

Second, according to the Perrian account, attitudes are three-place relations between subjects, absolute propositions and ‘guises’ of de se attitudes where propositions and guises are both shareable. What is special about de se attitudes is that the propositional contents of our de se attitudes can be entertained by others, but they cannot be entertained by others under the guises under which we believe them. According to this theory, both Ann and Bill believe a proposition, which is true if and only if the bear is about to attack Ann, but the guise under which Ann believes the proposition is different from the guise under which Bill believes the same proposition.

Both theories can avoid the problem posed by the bear attack case, because they deny that Bill shares all the relevant attitudes that explain Ann's action. They do so by assuming a special status of one's de se attitudes (i.e. de se exceptionalism).

I will return to the exceptionalist solution to the alleged problem of de se attitudes in Section 7. In the following two sections, I will consider two alternative proposals, according to which the fact that Bill does not perform the same action type as Ann can be explained without abandoning the traditional view of propositional attitudes.

5. Cappelen and Dever's strategy

We have mentioned that the argument presented in Section 3 makes three assumptions – in the bear attack case:

(i)	$B_0$, $B_1$ and D are all the relevant attitudes that explain Ann's action.
(iv)	The action that is explained by Ann's holding $B_0$, $B_1$ and D is the action, ‘curling up into a ball’.
(vi)	There are no relevant differences between Ann and Bill due to which, other things are not equal between them.

The proposal that is in line with the one discussed by Cappelen and Dever (2013) rejects (iv). According to this proposal, the following claim is true:

(iv')

The action that is explained by Ann's holding $B_0$, $B_1$ and D is the action, ‘Ann curls up into a ball’.

(iv') is compatible with the fact that there are many other action types that Ann performs. It only claims that what is explained by Ann's holding $B_0$, $B_1$ and D is the action, ‘Ann curls up into a ball’ – i.e. there is a law-like connection between attitudes $B_0$, $B_1$, D and the action that Ann curls up into a ball.

From (iv'), Cappelen and Dever argue that we should also reject (vi). This is because:

(vi')

There is a relevant difference between Ann and Bill: the action, ‘Ann curls up into a ball’, is not available to Bill.

Now, if (iv') and (vi') are both true, then the claim that Ann and Bill share all the relevant attitudes is compatible with the claim that they act differently. This is because given (iv') and (i) – the claim that $B_0$,$B_1$ and D are all the relevant attitudes, Explanation imply that:

(v')	If other things being equal, then Bill will perform the action, ‘Ann curls up into a ball’, in the bear attack case.

But given (vi'), the antecedent of (v') is false: Bill is not able to perform the action that Ann curls up into a ball. This means that Explanation will not predict that Bill will perform the same type of action that Ann performs.

In a word, Cappelen and Dever's strategy is trying to argue that the relevant action in question is the one that Ann curls up into a ball, and Ann, not Bill, performs this action, because the action is only available to Ann. Therefore, if Cappelen and Dever's strategy is right, the alleged problem of de se attitudes is avoided without rejecting the traditional view of propositional attitudes.

For the rest of this section, I am going to consider two possible critiques of Cappelen and Dever's strategy. The first objection is from Valente (2018), which I do not think is strong enough. Valente (2018) rejects the plausibility of (iv'). According to him, we should interpret (iv') as saying that Ann's action that is explained by the fact that Ann holds $B_0$, $B_1$ and D is an agent-specific action. An action is agent specific, if its description contains a particular agent. Examples of agent-specific actions are the action that Ann curls up into a ball, the action that Bill shouts from a distance, and so on. The actions, ‘curling into a ball’ and ‘shouting from a distance’ are agent-neutral actions. Their descriptions do not involve a particular agent. Valente argues that (iv') is plausible, only if a more general claim like the following is plausible:

(iv∗)

In general, the action types, which are involved in Explanation, should be agent specific.13

However, (${\mathrm{iv}}^{\ast }$) does not seem correct. Suppose that (${\mathrm{iv}}^{\ast }$) is true. Then Explanation would never predict that when two different agents share all the relevant attitudes, they will perform the same action. This would undermine the plausibility of Explanation itself. To see this point, consider the following case adapted from a case discussed by Valente (2018):

Nora Case.

Ella believes that Nora is in danger and that if someone calls the police, Nora will be saved. Since Ella wants Nora to be saved, Ella calls the police. Fiona also believes that Nora is in danger and that if someone calls the police, Nora will be saved. Fiona also wants Nora to be saved. Fiona does not know that Ella has already called the police. Valente (cf. 2018, 709)

In this case, Fiona seems to share all the relevant attitudes that motivate Ella to call the police. An intuitive judgement is that Fiona will perform the same action as Ella does.

However, if (${\mathrm{iv}}^{\ast }$) is true, then Explanation would make no such a prediction. To see this, suppose that (${\mathrm{iv}}^{\ast }$) is true. Then, the action that is explained by Ella's holding the relevant attitudes is the agent-specific action, ‘Ella calls the police’.¹⁴ But since Fiona is not able to perform the action that Ella calls the police, the ‘other things being equal’ clause in Explanation is false, and, therefore, Explanation does not predict that Fiona will perform any action. However, Explanation should predict that in the Nora case, the two agents will perform the same actions. Explanation is meant to capture our intuition that there is a law-like connection between attitudes and the action that they explain. Ella's de se attitudes in the case of Nora do not seem to play a role in explaining her action. So, the case is not a problematic case like the bear attack one. Explanation should be able to predict that Fiona will also call the police. This means that (${\mathrm{iv}}^{\ast }$) is false. But if (${\mathrm{iv}}^{\ast }$) is false, (${\mathrm{iv}}^{\prime }$) is false too.

My reply. I think Valente is wrong when he claims that Cappelen and Dever propose that the action, which is explained by the fact that Ann holds $B_0$, $B_1$ and D, is an agent-specific action. Rather, I think what Cappelen and Dever actually propose is that the explained action is an action that involves a particular bodily movement of Ann's, and this action is not open to anyone other than Ann herself.¹⁵

We can see this point in a clearer way by considering an alternative but equivalent description of Cappelen and Dever's strategy: in the bear attack case, the action, which is explained by Ann's holding $B_0$, $B_1$ and D, is the action, ‘bringing it about that Ann curls up into a ball’. The action involves a particular bodily movement of Ann's. However, it is still an agent-neutral action: the description does not specify who the agent is. This action is not open to Bill.

In line of this interpretation, Cappelen and Dever would say that in the case of Nora, what Ella does is the action, ‘bringing it about that someone calls the police’. This action is agent-neutral, and it is open to Fiona. Therefore, since Fiona believes and desires what Ella believes and desires, Explanation will predict that Fiona will also bring it about that someone calls the police.

The second objection: I think there is a more serious problem for Cappelen and Dever's strategy. The problem is that assumption (vi') seems wrong. To repeat:

(vi')

There is a relevant difference between Ann and Bill: the action type, ‘Ann curls up into a ball’, is not available to Bill.

It is tempting to claim that the action, ‘Ann curls up into a ball’ is not available to Bill. One reason for it might be that this action seems to be an action that is performed by a particular agent, and Bill is not the agent that is mentioned in the description of the action. But as I just said, the right way to interpret Cappelen and Dever's strategy is to claim that the description of the action is not about a particular agent but about a particular bodily movement of a particular individual's. But if so, then it does not seem clear that Bill cannot perform this action. Bill seems to be able to perform this action in many different ways. For example, Bill can bring it about that Ann curls up into a ball by using Bill's own hands to adjust Ann's body position. He can also try to bring it about that Ann curls up into a ball by suggesting, commanding or threatening Ann to do so. It is just that the action that Ann curls up into a ball involves Ann's bodily movement, over which Bill does not have direct control. It is not very easy for Bill to perform the action successfully. For example, Ann could disobey Bill when Bill commands Ann to curl up into a ball.

The point here can be illustrated more vividly by considering the following case:

Super Bill.

Ann and Super Bill are walking in the woods. Super Bill, unlike Normal Bill, can perform the action that Ann curls up into a ball very easily, because he has magical powers and is able to cast a spell that brings it about that Ann curls up into a ball (and he knows that he has the magical powers). Ann and Super Bill both realise that a bear is about to attack Ann; they both want the bear to leave. They act differently (at the same time): Ann curls up into a ball and Super Bill shouts at the bear from a distance.

This case has the same structure as the original bear attack case. The only difference is that Super Bill, unlike normal Bill, can perform easily the action that Ann curls up into a ball (or, if you like, the action ‘bringing it about that Ann curls up into a ball’). Then, given the assumption that $B_0$, $B_1$ and D are all the relevant attitudes that motivate Ann to perform the action that Ann curls up into a ball, the traditional view of propositional attitudes and Explanation will imply that if other things being equal, Super Bill will perform the action that Ann curls up into a ball.

But, now, Cappelen and Dever's strategy cannot be extended to the Super Bill case by saying that the ‘other things being equal’ clause is false. Their strategy fails to explain why Super Bill does not perform the action that Ann curls up into a ball. I think to explain Super Bill case, Cappelen and Dever will have to endorse the best action model I am going to develop in the next section.

Presumably, the alleged problem created by the original bear attack case is replicated by the case of Super Bill, and it is clear that Cappelen and Dever's strategy fails when it comes to the Super Bill case. So, if you don't like the idea that in the original case, normal Bill can perform the action that Ann curls up into a ball, then just focus on the Super Bill case.

Although Cappelen and Dever's strategy fails, I think it contains a right insight. The insight is that the action that is explained by Ann's holding the relevant action in the bear attack case is an action that involves a particular bodily movement of Ann's. My strategy, which I will develop in the next section, will preserve this insight. Let's now turn to my strategy.

6. My strategy

In this section, I am going to develop my own strategy. As I said, my strategy will maintain the key insight of Cappelen and Dever's strategy, according to which assumption (iv) is false –

(iv)	The action that is explained by Ann's holding $B_0$, $B_1$ and D is the action, ‘curling up into a ball’.

Instead, this insight says that the action in question is an action that involves a particular bodily movement of Ann's. This action is the action, ‘Ann curls up into a ball’ (or, if you like, the action, ‘bringing it about that Ann curls up into a ball’).

My strategy claims that Bill does not perform the action that Ann curls up into a ball because Bill does not think that this action is the best action. Therefore, my strategy also rejects assumption (i):

(i)	$B_0$, $B_1$ and D are all the relevant attitudes that explain Ann's action.

But before explaining how my strategy works, I will first propose a promising model of action explanation which will serve as a foundation. The model of action explanation that I propose is as follows:

• Best Action Model. Suppose that an agent x wants a proposition p to be true and believes that performing an action ϕ can bring it about that p is true. Then, x will perform ϕ, if:

  	(a)	x can perform ϕ,
  	(b)	x thinks that ϕ is the best action.

(In this formulation, I say that an agent x will perform an action ϕ, if x thinks ϕ is the best action. What type of attitude is the attitude of thinking that an action is the best? I want to keep neutral on this issue. The attitude might be the attitude of belief, preference, and so on.)

It is difficult to give a complete description of the notion of best action, but for our purposes, the following should suffice: action ϕ' is better than action ϕ, if –

• performing ϕ and performing ϕ' achieve the same desirable result(s), but performing ϕ' is easier or more efficient; or

• performing ϕ and performing ϕ' are equally easy and equally efficient, but performing ϕ' will achieve more desirable results than performing ϕ; or

• performing ϕ and performing ϕ' are equally easy and equally efficient, but performing ϕ will create more undesirable results.

An action ϕ is the best if there are no other actions such that they are better than ϕ.

I take it that the Best Action Model is very intuitive. Agents always perform what they consider to be the best actions. Later, I will show that Cappelen and Dever and exceptionalists will both have to accept this model as well.

Now, come back to the bear attack case. As I said, the action in question (i.e. the one that is explained by Ann's holding $B_0$, $B_1$ and D) is the action that Ann curls up into a ball. Given the Best Action Model, assumption (i), which says that $B_0$, $B_1$ and D are all the relevant attitudes, is false. The relevant attitudes are the following, which include two additional attitudes – the attitudes $B_2$ and A:

• $B_0$: the belief that Ann could express by saying, ‘The bear is about to attack me’;

• $B_1$: the belief that Ann could express by saying, ‘If I curl up into a ball when the bear is about to attack me, then the bear will leave’;

• $B_2$: the belief that Ann could express by saying, ‘I could bring it about that I curl up into a ball’;

• A: the attitude that Ann could express by saying, ‘The action that I curl up into a ball is the best action’;

• D: the desire that Ann could report by saying, ‘I want the bear to leave’.

It seems that in the bear attack case, Bill does not hold the attitude A: he does not think that the action that Ann curls up into a ball is the best action; rather, he thinks that it is better to perform the action that someone shouts from a distance.

Why does Bill think that the action that someone shouts from a distance is a better action than the action that Ann curls up into a ball? Well, from the short description of the bear attack case, we do not know the answer. The description does not provide enough information for us to say anything about this issue. Maybe Bill thinks that it is rather difficult to perform this action successfully, or he believes that if he tries to perform the action that Ann curls up into a ball, then the bear will start attacking him, which is an undesirable result that his performing the action that someone shouts from a distance can avoid. Or, maybe he believes that the action that someone shouts from a distance is the most efficient way to make the bear leave. These possibilities are all compatible with what is described in the bear attack case.

Now, if in the bear attack case, Bill does not hold A, then (iii), which says that Bill in this case shares all the relevant attitudes that motivate Ann to perform the action that Ann curls up into a ball is false. But if (iii) is false, Explanation does not imply that Bill in this case will perform the same action as Ann does – that Bill will bring it about that Ann curls up into a ball. The bear attack case does not pose any problem for the traditional view of propositional attitudes.

Moreover, with the Best Action Model, we can predict something more positive. We could attribute the following attitudes to Bill in the bear attack case, given the traditional view of propositional attitudes:

• $B_0$: the belief that Bill could express by saying, ‘The bear is about to attack you [addressing Ann]’;

• $B_3$: the belief that Bill could express by saying, ‘If someone shouts from a distance when the bear is about to attack you [addressing Ann], then the bear will leave’;

• $B_4$: the belief that Bill could express by saying, ‘I could bring it about that someone shouts from a distance’;

• $A^{\prime }$: the attitude that Bill could express by saying, ‘The action that someone shouts from a distance is better than the action that Ann curls up into a ball and any other possible action’;

• D: the desire that Bill could report by saying, ‘I want the bear to leave’.

Then, by the Best Action Model, Bill will perform the action that someone shouts from a distance.

I mentioned in the last section that the case of Super Bill creates the same problem as the original bear attack case does, and that Cappelen and Dever's strategy cannot explain it. Using the Best Action Model, we can handle the case easily. I think Cappelen and Dever will have to appeal to the Best Action Model eventually. In the case of Super Bill, Super Bill can perform the action that Ann curls up into a ball in a more easy way than Normal Bill, but he still does not perform this action as Ann does. The explanation my strategy offers is that: unlike Ann, Super Bill, for some reasons that are not specified in the description of the case, does not think that the action that Ann curls up into a ball is the best action. Like the original bear attack case, the Super Bill case does not create any problem for the traditional view of propositional attitudes.

One might still think that the best action explanation of the bear attack case or the Super Bill case does not sound right. I think this is mainly due to the fact that one could get misled easily by the bear attack case. To better appreciate my strategy, my suggestion is that we consider other similar cases where the action in question does not involve a particular bodily movement of a particular individual's. One example is the following:

Bear Shooting.

Anna and Billy are walking in the woods. Each of them has a gun with them. They both realise that a bear is starting to chase Ann. They however act differently. Anna shoots the bear; Billy shouts from a distance.

In this case, both Anna and Billy believe that Anna is about to be attacked by the bear. Both believe that if the bear gets shot, the bear will leave. Both believe that if someone shouts from a distance, the bear will leave. Both want the bear to leave. According to the traditional view of propositional attitudes, they believe and desire the same propositions. But they do not perform the same action.

The Best Action Model can explain the fact why Anna and Billy act differently in an intuitive way. According to it, Anna and Billy act differently, because they consider different actions to be the best. Anna thinks that the best action is the one that someone shoots the bear; Billy thinks that the best action is the that someone shouts from a distance. If Billy considered the action that someone shoots the bear to be the best, then he would have performed this action. This explanation of the bear shooting case comes more naturally than the explanation that I offer to the original bear attack case. This is because the action that is explained by Anna's holding the relevant attitudes does not involve Anna's particular bodily movement. The action that the bear gets shot is equally open to Anna and Billy.

Now, the two cases – the original bear attack case and the bear shooting case – are similar in structure. They should have a unified explanation. My strategy offers a unified explanation. If I am right, neither of them poses a problem for the traditional view of propositional attitudes.

Let's consider one possible objection against my strategy. It goes as follows: attitude A is the attitude that Ann can express by saying, ‘The action that I curl up into a ball is the best action for me’. According to the traditional view of propositional attitudes, for Bill to hold A is for Bill to hold an attitude that he can express by saying, ‘The action that you curl up into a ball is the best action for you’. Now, in the bear attack case, it is not implausible to attribute this attitude to Bill. Then, it is still puzzling why Bill act differently from Ann.

My reply: The objection depends on a particular analysis of attitude A. According to this analysis, A is one of Ann's de se attitude. However, this analysis does not seem right. Here is why: it is natural to say that in the bear attack case, Ann and Bill disagree about which action is the best. However, if the attitude A is to be analysed as one of Ann's de se attitudes, then there is no disagreement between Ann and Bill – they both agree that the action that Ann curls up into a ball is the best action for Ann (and the action that someone shouts from a distance is the best for Bill).

Therefore, the de se analysis of the attitude A is wrong. But I have to admit that I do not have a positive account of the attitude in question. All I want to claim here is that the two agents in the bear attack case disagree about which action is the best. The traditional view might not be able to account for the attitude in question. But even if this is true, the problem here has nothing to do with de se attitudes.

To sum up, the Best Action Model fares well with the bear attack case (as well as the Super Bill case). If my strategy works, then the bear attack case is not philosophically interesting. The case does not pose any problem for the traditional view of propositional attitudes, nor does it illustrate anything interesting about de se attitudes.

7. Comparison

I have proposed the Best Action Model and argued that the bear attack case does not pose any problem for the traditional view of propositional attitudes. Therefore, the argument for de se exceptionalism based on the bear attack case is undermined.

In this section, let's consider the same issue from a different perspective. In the bear attack case, Ann and Bill perform different actions. Let's take this as a data point. I am going to show that my strategy can explain this data point in a more elegant way than the strategies of the exceptionalist and Cappelen and Dever

First, exceptionalists claim that Ann and Bill act differently because they hold different beliefs when they believe that the bear is about to attack Ann. This cannot be accommodated by the traditional view of propositional attitudes. They think that we need the centred-worlds account (or any other revisionary theories of de se attitudes). For example, according to the centred-worlds view, Ann and Bill perform different actions, because Ann and Bill believe different centred-worlds propositions. Only Ann believes the proposition {$⟨w,x⟩$: the bear is about to attack x in w}; only Bill believes the proposition {$⟨w,x⟩$: the bear is about to attack the addressee of x in w}.

Second, Cappelen and Dever's strategy explains the data point by appealing to two assumptions: first, the relevant action type is the action, ‘Ann curls up into a ball’ (or the action, ‘bringing it about that Ann curls up into a ball’); second, the relevant action type is not available to Bill. Therefore, Bill does not perform the same action type as Ann.

Finally, my strategy explains the data point by appealing to the idea of the best action. According to my strategy, Ann and Bill in the bear attack case act differently because they disagree about which action is the best: Ann thinks that the action, ‘Ann curls up into a ball’ is the best; Bill thinks that the action, ‘someone shouts from a distance’, is the best action.

My strategy is better than the other two. First, it is better than Cappelen and Dever's strategy. It explains the case by appealing to the doubtful assumption that what Ann does is not open to Bill. Besides, even if the assumption is true, it does not have anything to say about why in the case of Super Bill, Ann and Super Bill perform different actions. My strategy can explain both cases by appealing to the Best Action Model. It has more explanatory power than Cappelen and Dever's strategy.

Second, my strategy is better than the exceptionalist strategy. This is because while exceptionalist strategy has to appeal to a revisionary theory of propositional attitudes to explain the bear attack case (and the case of Super Bill, and the case of bear shooting), my strategy can just stick to the traditional view of propositional attitudes. It is more conservative than the exceptionalist strategy.

In addition, my strategy appeals to the Best Action Model, which I have mentioned is a very intuitive model. I am going to argue that the centred-worlds theorist has to appeal to this model as well to explain some other cases. If that is true, this means that, first, my strategy uses the resource that exceptionalists should use anyway; second, my strategy is more unified than the exceptionalist strategy (at least, more unified than the exceptionalist strategy that accepts the centred-worlds account).

Consider the following case:

BearAttack∗

Ann and Bill are walking in the woods when a bear starts following them. Ann and Bill both believe that the bear is about to attack both of them. They both want the bear to leave. They act differently. Ann curls up into a ball; Bill climbs a tree.

Suppose that the centred-worlds account is true. Then it seems that Ann and Bill in the new bear attack case both hold the following attitudes:

• $B_0^{\ast }$: the belief that both Ann and Bill could express by saying, ‘The bear is about to attack me’ – the belief whose content is the centred-worlds proposition {$⟨w,x⟩$: the bear is about to attack x in w};

• $B_1^{\ast }$: the belief that both Ann and Bill could express by saying, ‘If I curl up into a ball (when the bear is about to attack me, then the bear will leave)’ – the belief whose content is the centred-worlds proposition {$⟨w,x⟩$: If x curls up into a ball, then the bear will leave in w};

• $B_2^{\ast }$: the belief that both Ann and Bill could express by saying, ‘If I climb a tree (when the bear is about to attack me), then the bear will leave’ – the belief whose content is the centred-worlds proposition {$⟨w,x⟩$: If x climbs a tree, then the bear will leave in w}.

• D: the belief that both Ann and Bill could report by saying, ‘I want the bear to leave’ – the desire whose content is the centred-worlds proposition {$⟨w,x⟩$: The bear leaves in w}.

Suppose that $B_0^{\ast }$, $B_1^{\ast }$ and D are all the relevant attitudes that motivate Ann to curl up into a ball in the new bear attack case, and that $B_0^{\ast }$, $B_2^{\ast }$ and D are all the relevant attitudes that motivate Bill to climb a tree in this case. Then, we can claim that Ann and Bill in this case hold all the same attitudes. By Explanation, we can infer that Ann and Bill in this case will perform the same action. This contradicts with the fact that Ann and Bill in this case act differently. The centred-worlds account is in trouble.

There are two possible ways out for the centred-worlds theorist. First, she can adopt Cappelen and Dever's strategy and claim that Ann's action involves a bodily movement of Ann's, which is not open to Bill, and Bill's action involves a bodily movement of Bill's, which is not open to Ann. As I have argued, this strategy is not promising. The second strategy that the centred-worlds theorist can appeal to is mine. She can claim that Ann and Bill perform different actions, because they think different actions are the best. In particular, Ann thinks that the action, ‘curling up into a ball’, is the best action; Bill thinks that the action, ‘climbing a tree’, is the best action. This strategy is intuitively correct.

In a word, to avoid the problem raised by the new bear attack case, the best strategy for the centred-worlds theorist is to appeal to the Best Action Model. But if we have the Best Action Model in hand, we can explain the original bear attack case without revising the traditional view of propositional attitudes.

8. Conclusion

In this paper, I have argued that the argument for de se exceptionalism based on the bear attack case is not persuasive. The case can be explained by appealing to the Best Action Model.

If both the bear attack case and de se ignorance cases do not lend any support to de se exceptionalism, we should probably reject it. This is an interesting conclusion. The centred-worlds account of de se attitudes has been widely accepted in both philosophy and linguistics, but it assumes de se exceptionalism. If de se exceptionalism is not well motivated, the centred-worlds view is problematic. We need to reconsider how to characterise de se attitudes.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Funding

This work was supported by China Scholarship Council – University of St Andrews Scholarships [201806360003].

Additional information

Funding

This work was supported by China Scholarship Council – University of St Andrews Scholarships [201806360003].

Notes

1 Perry (1977, 1979) and Lewis (1979). Also see Castañeda (1966).

2 Cf. Boer and Lycan (1980), Stalnaker (1981), Millikan (1990), Tiffany (2000), Spencer (2007), Cappelen and Dever (2013) and Magidor (2015).

3 Cf. Ninan (2016, 2020), Valente (2018), Torre (2018) and Torre and Weber (2021).

4 Exceptionalists also try to show that once we revise the traditional notion of de se attitudes, the problem is not replicable by non-de se attitudes; so, unlike the problem raised by de se ignorance case, the problem here is specific to de se attitudes. I am not going to elaborate on this point in this paper because I think that the bear attack case does not pose a problem for our traditional notion of propositional attitudes in the first place.

5 The term is sometimes used in a broader way such that it is interchangeable with ‘indexical attitudes’ or ‘self-locating attitudes’, attitudes that could usually be expressed or self-reported by indexicals (e.g. ‘I’, ‘now’, and ‘here’). For simplicity, I will only focus on first-person attitudes.

6 This assumption is debatable but widely accepted. See, for example, Fodor (1994) and Braun (2001).

7 Is $B_1$ a de se belief of Ann's? The belief can be expressed by Ann's using ‘I’ and ‘me’. So, it seems that it is Ann's de se belief. However, I think that the first-person pronouns here do not function as first-person pronouns. The belief can also be expressed without using any first-person pronoun. For example, Ann can express $B_1$ by saying that if anyone who is about to be attacked by the bear curl up into a ball, then the bear will leave. When I characterised the notion of de se attitude, I said they are attitudes that can usually be expressed by their holders using first-person pronouns. I did not intend to give a definition of de se attitude. I just offered a rough heuristic for detecting them. It is possible that by the end of the day, we do not need the heuristic; rather, the notion of de se attitude is in light of theorising (cf. Ninan 2016, 90).

8 Here I assume that $B_0$ could be expressed by Bill using some linguistic expressions. The following discussion does not hinge on this assumption – all we need here is the claim that $B_0$ has that absolute truth condition.

9 One might claim that Bill does not curl up into a ball because he believes that the bear is about to attack Ann not him. This reply is not promising, for Ann also believes that the bear is about to attack her not Bill, but she curls up into a ball. In the traditional view of propositional attitudes, what they believe is the same proposition. Therefore, in the traditional view, what explains Bill's not performing the action cannot be what he believes.

10 The centred-worlds account is inspired by Lewis (1979) (also see Chisholm 1981). The account is later developed by many authors such as Egan (2007), Ninan (2012) and Kindermann (2019).

11 The terminology of ‘boring’ and ‘interesting’ centred-worlds propositions follows Egan (2007).

12 The idea of the Fregean account is discussed by Frege (1918/1956); the Perrian account is discussed by Perry (1977).

13 Put it differently, in general, the law-like connection should be between attitudes and agent-specific action types.

14 In other words, there is a law-like connection between Ella's attitudes and the action type, ‘Ella calls the police’.

15 If this is not what they actually propose, then they should have accepted my interpretation in order to avoid Valente's objection.