• No results found

Difficult Questions and Cognitive Limitations

Chapter 5: Permissivism, Cognitive Capacities and Limitations Background

5.2 Difficult Questions and Cognitive Limitations

Simpson’s account of how different cognitive capacities make appropriate different epistemic standards ultimately boils down to a variation of the claim that people can rationally disagree because they have limited cognitive resources or capacities. This claim is made explicitly by Gaus (2011) and Rosa (2012; 2016), and is also implicit in Rawls’s (1993) account of the first, third and fifth burdens of judgment.

Recall that according the first burden of judgment, people can rationally disagree because “the evidence… bearing on the case is conflicting and complex and thus hard to assess and evaluate” (Rawls 1993: 56). Similarly, according to the fifth burden, “there are different kinds of normative considerations of different force on both sides of an issue and it is difficult to make an overall assessment” (ibid: 57). As I argued in Chapter 2, both the first, third and fifth burdens claim that because it is difficult to evaluate the overall normative force of one’s evidential or moral reasons, disagreement about these issues, even given the same total body of evidence, can be rational.

My claim, in this section, is that disagreement that comes about because the overall normative force of our reasons is difficult to assess would be rational only if agents should lower their epistemic standards because of their cognitive limitations. While Rawls himself says little on this topic, Gaus’s (2011) account of reasonable disagreement provides a plausible account linking the difficulty of an issue to the rationality of disagreement about that issue. Gaus’s argument consists of three claims.

97 A) People have limited cognitive resources.

B) Avoiding errors in reasoning can in some cases require large amounts of cognitive resources.

C) There is a point beyond which people are not obligated to use their cognitive resources to avoid errors in reasoning.

These three claims together imply that when matters are difficult to assess, disagreement can be reasonable. In order to show why this is the case, I shall elaborate on each claim in turn.

Gaus’s first claim is that people have limited cognitive resources. Reasoning requires time and effort and people only have a finite amount of time and effort that they can spend on reasoning. People cannot continue reasoning indefinitely; they eventually have to act on the beliefs that they have reasoned about so far (ibid: 246- 247). Gaus’s account is concerned with how much reasoning an agent does, that is, how much of an agent’s system of beliefs or evidence she looks through to identify supporting and defeating considerations. Apart from identifying a consideration, we might suppose that cognitive resources are also expended in assessing the strength of that consideration. It takes at least some effort for an agent to make herself aware of her biases and screen them off when assessing how significant a given piece of evidence is. We may also think that correspondingly that the amount of effort a person can continuously exert over a given period of time is limited as well. In addition to the claim that cognitive resources are limited, Gaus’s second claim is that avoiding errors in reasoning can consume large amounts of cognitive resources. Consider that when we casually examine a piece of evidence, we are prone to hastily over or underestimate the strength of that consideration. For instance, as Kornblith notes, we are often prone to drawing conclusions about a population on the basis of a very limited sample size. He points to the common tendency to draw conclusions about consumer products on the basis of a single person’s bad experience (Kornblith 1983: 42-44)39. This tendency seems to indicate that refraining from hastily forming beliefs on inadequate inductive evidence takes more effort than many among us are willing to expend. Gaus notes that the amount of reasoning one has to engage in so as to take into account all the relevant considerations found in an agent’s system of beliefs exceeds the capabilities of ordinary reasoners. He says that “only ideally rational agents, capable of scanning

39

To be sure, Kornblith (1983) does try to argue that such inferences are justified in situations where we cannot obtain better evidence. However, he notes that these inferences still run afoul of what he calls rules of ideal reasoning.

98

their entire belief system for lurking defeaters” could completely avoid errors in reasoning (Gaus 2011: 146). We need not agree with Gaus in thinking that every question will require us to scan through our entire system of beliefs in order to avoid errors in reasoning. Perhaps in particularly simpler cases, it is fairly obvious what the evidence regarding the proposition can be, and scanning through this set of beliefs is within the grasp of most ordinary agents. For instance, splitting the dinner bill equally between two persons is something that most ordinary agents are capable of and does not require anyone to scan through their entire system of beliefs. Other questions might require more reasoning and this would mean that the agent had to scan more of her system of beliefs to identify the relevant considerations. Summing up, reasoning well requires cognitive resources, and cases which require more reasoning or more careful reasoning require more cognitive resources. Given that people have limited cognitive resources, questions which use up most of if not more cognitive resources than a person may have available will be difficult to assess without committing some kind of error in reasoning. Therefore, the normative force of an agent’s reasons would be difficult to assess only because of her cognitive limitations.

If the difficulty of assessing the normative force of one’s reasons can render the resulting disagreement rational, this can only be because people, due to their cognitive limitations, ought to accept epistemic standards which are constituted by relatively less demanding epistemic norms. If agents could not permissibly lower their standards in this way, then, they ought to reason according to the more demanding epistemic norms even though they are cognitively limited. If so, the fact that they are cognitively limited and the reasons are therefore difficult to assess would play no role in making any resulting disagreement rational. Therefore, in order for the first and fifth burdens to be sources of rational disagreement, it must be permissible for agents to employ less demanding epistemic standards in response to their cognitive limitations.

This must similarly be the case for the third burden of judgment, according to which people rationally disagree because their concepts to some degree or another are vague and subject to borderline cases. Williamson explains the existence of vague predicates by appealing to cognitive limitations that prevent us from clearly identifying whether a given borderline case falls under the predicate. As I suggested in Chapter 2, these cognitive limitations are of the same sort that could make inferences when the evidence is conflicting and complex subject to rational disagreement. If so, it must also be the case that in order for the third burden of

99

judgment to generate rational disagreement, it must be permissible for agents to employ less demanding epistemic standards in response to their cognitive limitations.

As mentioned earlier, cognitive limitations also play a crucial role in Simpson’s account. Indeed in reference to Detective, Nancy Drew is unable to reliably assess the sincerity of testimony and Veronica Mars similarly cannot assess the evidence as a whole, and has to proceed piece by piece. Their disagreement is rational only if it is permissible for each to employ epistemic standards which are less demanding in ways that correspond to those respects in which they are cognitively limited. 5.3 Limitation of Epistemic Obligations by Cognitive Capacity

It is precisely this last claim which is defended by Gaus. According to the third claim made by Gaus, there is some point beyond which people are not obligated to use their cognitive resources to avoid errors in reasoning. Gaus, for instance, endorses the following account of when individuals have sufficient reason to hold a particular doxastic attitude. Let’s call this principle Provisionally Sufficient Reasons:

Provisionally Sufficient Reasons: “Alf has (provisionally) a sufficient reason R40 if and only if a ‘respectable amount’ of good reasoning by Alf would conclude that R is an undefeated reason” (ibid: 250)

Provisionally Sufficient Reasons implies that people do not need to engage in a more than respectable amount of reasoning in order to rationally hold a particular doxastic attitude towards a proposition. After all, if a given, “respectable” amount of reasoning is sufficient for the resulting doxastic attitude to be rational, then a more than sufficient amount of reasoning cannot be obligatory. We might, in addition, think that apart from the amount of reasoning, there could also be some “respectable” amount of effort we put into reasoning beyond which there is no further obligation to ensure that we do not under or over-estimate the strength of a given piece of evidence.

There are different ways in which we could determine what level of cognitive resources we have to commit in order to have engaged in a respectable amount and quality of reasoning. For instance, it could be that people have an obligation to use all their available cognitive resources. If people have different amounts of cognitive resources available, what counts as a respectable amount of reasoning

40

In case it is not already clear, I will take the elocutions “A has sufficient reason to believe P” and “It would be rational for A to believe P” to be interchangeable.

100

would differ between them. Or, it could be that people ought to expend some fixed fraction of their total cognitive resources in reasoning, but no obligation to expend more than that. Alternatively, there could be some fixed amount of resources that people are obliged to expend. People who have very little in the way of cognitive resources would be obliged to expend all or most of their cognitive resources while those who are much wealthier in terms of cognitive resources need to spend only a small fraction of their resources before they have satisfied their epistemic obligations. I shall discuss these different accounts later in this chapter. Suffice it to say that for any of these accounts, it will sometimes be the case that a more than the obligatory amount of cognitive resources needs to be expended in order to avoid errors in reasoning.

Suppose that in a given case, a more than obligatory amount of resources is necessary in order to avoid erroneous reasoning. Alfred and Betty share a body of evidence which contain two pieces of evidence, R1 and R2 which pertain to whether P. Suppose that after a “respectable” amount of reasoning, Alfred uncovers within his evidence R1 which supports a belief in P, but not R2. Betty, who shares the same body of evidence uncovers after a “respectable” amount of reasoning R2 which supports disbelief in P. Since the amount of reasoning required to account for all the considerations, R1 and R2, relevant to P exceeds the amount of reasoning that Alfred and Betty are obligated to put in when considering whether P, we can treat this as being a case in which the question of whether P is difficult to assess. Moreover, since Alfred and Betty have both put in a respectable amount of reasoning, according to Provisionally Sufficient Reasons, they would have sufficient reason and are therefore rational in their corresponding beliefs. This example should make clear at least one way in which a proposition being difficult to assess creates room for rational disagreement. We can generate other examples whereby the amount of cognitive resources required to completely avoid errors in reasoning exceeds the amount which agents are obligated to expend in reasoning. In any such case, the fact that the cognitive cost of completely avoiding errors is so high, combined with the supposed fact that people may permissibly expend fewer resources, makes it the case that rational disagreement is possible.

It is difficult to dispute that the cognitive costs in completely avoiding errors in reasoning can be very high in some cases. The more questionable assumption in a view like Gaus’s is that it is sometimes epistemically permissible to make errors while reasoning about a proposition. In order to argue that difficulty in assessing whether a proposition is true does not give rise to rational disagreement, I submit

101

that making an error in reasoning is always irrational. In addition, I shall argue that it is always an error to believe that P, when I have at least as strong a reason not to. Let us call the combination of these claims the Naïve account of epistemic rationality. We can contrast the Naïve account with what I call a Forgiving account according to which there are some cases in which I rationally hold a given doxastic attitude even if I make an error in reasoning or there are some cases where I do not make an error even though I believe that P even though I have at least as strong a reason not to. I shall first sketch out an argument for the Naïve account, followed by some objections to it. I shall subsequently reply to these objections to show that the Naïve account is in fact true.