1. Introduction
Rational or epistemically justified beliefs are often taken to be defeasible.Footnote 1 Take the case of perception. Many epistemologists argue that immediate perceptual beliefs provide us with justification in the absence of reasons to doubt them. For example, Pollock and Cruz state ‘if something looks red to you and you have no reason to think that it is not red then you are permitted to believe it is red’ (Reference Pollock and Cruz1999: 157). Similarly, Wedgwood thinks that you are permitted to believe that p when ‘you have an experience or apparent perception as of p’s being the case, and have no special reason to think that your experiences are unreliable in the circumstances’ (Reference Wedgwood2002: 276). This gives rise to the notion of prima facie justification. Beliefs that have some initial justification-conferring property, such as being based on sufficient evidence or formed via a reliable process are prima facie justified.
This defeasibility of prima facie justification is a widely acknowledged feature of both internalist accounts of justification, such as evidentialism, as well as externalist accounts of justification, such as process reliabilism.Footnote 2 Furthermore, recent developments in collective epistemology suggest that understanding the role of defeat and its relation to group evidence are crucial elements in any analysis of collective justification.Footnote 3
Let us call the doctrine that we can have prima facie justified beliefs which are defeasible, DefeatismFootnote 4 :
Defeatism: Doxastic attitudes can have the status of being prima facie justified. That is, some doxastic attitude D can be prima facie justified by having some justification-conferring property P while being defeated and, therefore, lacking ultima facie justification.
Defeatism, while not without its critics, is motivated by a plethora of so-called defeater casesFootnote 5 ; that is, cases in which it seems that some doxastic attitude D despite having the justification conferring property P is unjustified due to the existence of some defeater φ. Accordingly, many have argued that any theory of epistemic justification needs a so-called no-defeater clause that specifies the conditions under which prima facie justified beliefs are ultima facie justified.
Usually, defeater clauses come in one of two forms:
Negative No-Defeater Clause: S’s prima facie justified belief that p is ultima facie justified iff there is no defeater φ (that S should have had), that is defeating the belief’s prima facie justification.
Positive No-Defeater Clause: If S’s prima facie justified belief that p is ultima facie justified then S has a defender Φ, that is defending the belief that p from potential defeaters φ (that S should have had).
In this paper, I will survey various negative evidentialist as well as positive responsibilist no-defeater clauses and develop a general taxonomy of defeater cases against which these clauses can be tested.Footnote 6 Despite influential arguments that evidentialist understandings of justification are ill-equipped to handle the full spectrum of defeater cases, I will demonstrate that evidentialism has the right tools to make sense of all kinds of defeaters, including propositional and normative defeaters.Footnote 7 Yet, I will do so in an unconventional way. Namely, I will identify two desiderata underlying responsibilist treatments of defeater cases and argue that we can understand these desiderata in light of a positive higher-order evidentialist clause. This is an important result because it allows those sympathetic to evidentialist understandings of epistemic justification to retain an evidentialist theory while taking into account defeater intuitions.
In particular, I will defend the following evidentialist higher-order clause:
-
Higher-Order Evidence Clause <HOE>:
-
S belief that p is undefeated iff:
-
(a) S possesses some higher-order evidence EH, which is sufficient to support q, the proposition that the total relevant evidence ETOTAL on balance supports p and
-
(b) S’s belief that p is properly based on EH.
-
Before I proceed, let me note that, while higher-order in some to-be-defined sense, <HOE> does not over-intellectualise justification in any problematic way as traditional internalist higher-order belief requirements arguably do.Footnote 8 Instead, as I will argue below [§8], <HOE> is cognitively very undemanding if we understand the underlying terms ‘evidence’, ‘evidential support’ and ‘epistemic basing’ in the right way.
Here is the outline. I will start with some conceptual remarks about evidence [§2] and various types of defeaters [§3]. These preliminaries enable me to illustrate how conventional evidentialist strategies fail to give us the right verdict concerning the full range of defeater cases [§4]. Analysing the ways in which these conventional evidentialist strategies fail, pushes us towards a responsibilist understanding of defeat [§5], and helps me to develop a general taxonomy of defeater cases [§6]. While this proposed taxonomy suggests a responsibilist solution, I will show that epistemic responsibility is best understood via the above-stated higher-order evidentialist condition [§7] [§8]. This not only sparks hope for those who want to defend an evidentialist understanding of epistemic justification in general but also allows us to put forward a universal treatment of defeater cases which could be understood as bypassing the doctrine of Defeatism [§9].
2. Evidence
Before talking about defeaters, it is important to start with some preliminaries on evidence and its role in epistemic justification.
First, I will remain neutral concerning the nature of evidence. In other words, I will not take sides in the debate on whether pieces of evidence are mental states or propositions picked out by those mental states.Footnote 9 Relatedly, some argue that false propositions can be part of one’s evidence, while those who defend a factive understanding of evidence claim that only true propositions can constitute evidence. Following Williamson (Reference Williamson2000), factive understandings of evidence are increasingly popular. However, there is still a considerable number of epistemologists who understand evidence to be non-factive. Accordingly, I decided to organise the following discussions as if I relied on a non-factive understanding of evidence. Since if we can provide an evidentialist no-defeater clause relying on a non-factive understanding of evidence, we have done more than required for the evidentialists who have a factive understanding of evidence.
With these preliminaries about the nature of evidence in mind, we can now draw different distinctions between different types of evidence, such as possessed and unpossessed, accessible and inaccessible, or available and unavailable evidence.
Let us start with the evidential base, that is, the entire evidence possessed by an epistemic agent.
The evidential base (in short: E BASE ): The total evidence possessed by an epistemic agent S.
There are different ways to understand evidence possession. For example, mentalists might define EBASE to be the entirety of some relevant mental states, such as experiences, whereas perspectivalists hold that the evidence possessed by an epistemic agent is the evidence that they have a certain doxastic relationship with [§4].Footnote 10 This could, for example, be all of one’s justified beliefs (if evidence is non-propositional) or all of the propositions one knows (if evidence is propositional).
Depending on our understanding of evidence and evidence possession, we can make some further distinctions here. For example, we may distinguish between the accessible possessed evidence and the inaccessible possessed evidence, where accessibility roughly means that S could become aware of the evidence upon reflection.Footnote 11
The accessible base: The total evidence possessed by and accessible to S.
Based on this notion of accessibility, we can further distinguish between accessible evidence that has been accessed (that is, evidence that S was or is aware of) and evidence that is accessible but has not been accessed (that is, evidence that S could become aware of).
The accessed base: The total evidence which is accessible to and has been accessed by S.
Furthermore, we could consider some normative restrictions on the accessible evidence to differentiate between the part of the accessible base which we should have accessed from the part which we could blamelessly ignore. Let us call this the required base:
The required base: The total unaccessed evidence that S should have accessed.
So far, we have introduced the notion of an evidential base and distinguished various subparts of it. However, beyond the evidential base is a wider set of evidence pertinent to epistemic considerations. It includes evidence that is not possessed by S but is epistemically relevant for S in a looser sense. It includes all the propositions that S should know and/or could know given some effort, or equivalently, all the evidence that S could come to possess and/or should come to possess. I will call this the available evidence.
The available evidence: The total unpossessed evidence that S could come to possess.
The available evidence will be important in the following discussions of propositional and normative defeaters [§3]. One of the questions will be whether prima facie justified beliefs can be defeated only by unpossessed evidence that we should have possessed, or also by unpossessed evidence that we are not epistemically required to have but which is nonetheless available to S. That is, we could further differentiate between the unpossessed but required and unpossessed and not required evidence:
The required evidence: The total unpossessed and available evidence that S should have had.
Finally, if we take the union of the evidence possessed by S and the evidence the evidence available to S, we get what I call the total evidence:
The relevant evidence (in short: E TOTAL ): The total evidence that is relevant to S’s epistemic situation.
We can understand ETOTAL as the entire evidence that bears on the proposition in question and is relevant to the epistemic situation. As such, it includes the entire evidence possessed by S as well as all of the evidence that is available to S in some important sense.Footnote 12
In sum, we can distinguish between different subsets of the total evidence via the following descriptive relations possession, availability, and accessibility, as well as the two normative relations, should have accessed and should have possessed. This gives us the following tree diagram [Fig. 1].
Figure 1. Evidence.
Much of the following discussions will be centred around the questions of whether we can or should make all of these distinctions and whether some of these distinctions are extensionally equivalent. Regardless of how we answer these questions, having an intuitive understanding of these distinctions will turn out to be useful when characterising different kinds of defeaters [§3] and discussing extant no-defeater clauses [§4] [§5].
3. Defeaters
Defeaters come in various types, and the two central kinds are mental state defeaters and propositional defeaters.Footnote 13 In other words, defeaters can be true propositions external to the perspective of the epistemic agent (propositional defeaters), or conditions internal to and/or within the perspective of the epistemic agent (mental state defeaters) (Sudduth Reference Sudduth2008; §1).
Remaining largely neutral on the nature of evidence enables us to give a general evidential characterisation of defeaters that encompasses mental state defeaters, as well as propositional defeaters. That is, we can understand defeaters as evidence relevant to the epistemic situation of S that has some defeating force concerning some otherwise justified belief of S. Having this evidential understanding of defeat permits us to use the above-introduced relations (possessed, accessible, available, etc.) to neatly distinguish various kinds of propositional and mental state defeaters.
First, propositional defeaters are outside of S’s evidential base. That is, S’s belief that p is factually defeated by some unpossessed evidence φ if acquiring φ would result in a loss of justification of S’s belief that p (Sudduth Reference Sudduth2008: §2; Bergmann Reference Bergmann2005: 154). For example, my otherwise justified belief that there is a barn in front of me might be defeated by the true proposition φ that I am in Fake Barn County (Goldman Reference Goldman and Pappas1979). Accordingly, we can define propositional defeaters as follows:
Propositional Defeater: Some evidence φ is a propositional defeater iff, (i) φ has enough defeating force to render S’s prima facie justified belief that p unjustified and (ii), φ is unpossessed (i.e. outside EBASE).
These propositional defeaters are traditionally understood as defeaters for knowledge rather than justified beliefs. However, more recently, many epistemologists have argued that some specific kind of propositional defeaters, called normative defeaters, can also defeat or weaken the epistemic status of an otherwise justified belief.Footnote 14 If I should have known that I am in Fake Barn County, that is, if my ignorance with respect to φ is epistemically irresponsible (in a sense to be specified [§5]), φ is a normative defeater. Accordingly, we can characterise normative defeaters as ‘potential defeater[s] that the subject does not actually possess but should’ (Graham and Lyons Reference Graham, Lyons, Brown and Simion2021: 45) or ‘evidence she does not possess but should have possessed’ (Nottelmann Reference Nottelmann2021: 1183; see also Goldberg Reference Goldberg2016; Reference Goldberg2018: ch.6) or ‘a doubt or belief that S ought to have and that indicates that S’s belief that p is either false or unreliably formed or sustained’ (Lackey Reference Lackey2008: 45; see also Lackey Reference Lackey1999: 475; Lackey Reference Lackey2016: 366).
Accepting that there are normative defeaters raises further questions about the relationship between the evidence we should have possessed and its availability. In particular, we might ask ourselves whether ‘ought have possessed’ implies ‘could have possessed’ and/or whether ‘could have possessed’ implies ‘ought have possessed’. While some have suggested that all normative defeaters are defeaters that are available (Harman Reference Harman1980: 164; Goldberg Reference Goldberg2018: 191; see also Nottelmann Reference Nottelmann2021: 1186), or at least indicated by the available or possessed evidence (Lackey Reference Lackey2016: 366), the literature is less clear on whether all available defeaters are normative defeaters. In acknowledgement of this, I will assume that all normative defeaters are available defeaters but not vice versa. An unopened letter containing some important information concerning the whereabouts of my friend might be an available, an unavailable, or an available and normative defeater for my belief that she is in town. For example, if the letter is on my desk it might be available and normatively required, while it would only be available but not normatively required if it is at my colleague’s desk and neither if it were still at the post office.Footnote 15
Accordingly, we can define two types of propositional defeaters, available defeaters and normative defeaters, where the latter is a subtype of the former:
Available Defeaters: Some evidence φ is an available propositional defeater for S’s belief that p iff, (i) φ has enough defeating force to render S’s prima facie justified belief that p unjustified and (ii), φ is unpossessed, available but not required (i.e. outside EBASE but inside the required evidence).
Normative Defeater: Some evidence φ is a normative propositional defeater for S’s belief that p iff, (i) φ has enough defeating force to render S’s prima facie justified belief that p unjustified and (ii), φ is unpossessed, available and required (i.e. within the required evidence).
Second, mental state defeaters are within the evidential base EBASE of S. That is, a mental state defeater is some evidence possessed by S that has some putative defeating force with respect to some prima facie justified belief of S. Internalists characteristically deny that there are propositional or normative defeaters but both internalists and externalists typically agree that mental state defeaters can defeat otherwise justified beliefs.
Mental State Defeater: Some evidence φ is a mental state defeater for S’s belief that p iff, (i) φ has enough defeating force to render S’s prima facie justified belief that p unjustified and (ii), φ is possessed (i.e. φ is within EBASE).
Here, we could further distinguish between different kinds of mental state defeaters, including doxastic, reflective, and inaccessible mental state defeaters.Footnote 16 A doxastic defeater is some piece of defeating evidence φ of which one is aware (φ is within the accessed base). A reflective defeater is some piece of defeating evidence φ of which one is not aware but of which one could become aware upon reflection (φ is within the accessible base but outside the accessed base). An inaccessible defeater is some piece of defeating evidence φ which is possessed by S but which S cannot become aware of upon reflection (φ is outside the accessible base).
Doxastic Defeater: Some evidence φ is a doxastic defeater for S’s belief that p iff, (i) φ has enough defeating force to render S’s prima facie justified belief that p unjustified and (ii), φ is possessed and accessed (i.e. φ is within the accessed base).
Reflective Defeater: Some evidence φ is a reflective defeater for S’s belief that p iff, (i) φ has enough defeating force to render S’s prima facie justified belief that p unjustified and (ii), φ is possessed and accessible but not accessed (i.e. φ is within the accessible base and outside the accessed base).
Inaccessible Defeater: Some evidence φ is an inaccessible defeater for S’s belief that p iff, (i) φ has enough defeating force to render S’s prima facie justified belief that p unjustified and (ii), φ is possessed and inaccessible (i.e. φ is within EBASE and outside the accessible base).
Whether these distinctions between doxastic, reflective, and inaccessible defeaters are important, or whether they even make sense depends on various background assumptions about the nature of evidence and evidence possession, as well as assumptions about how defeaters exert their defeating force. For example, we could ask, following Alston (Reference Alston1988), whether accessible defeaters need to be accessible in the sense that we can access the evidence that constitutes the defeater or accessible in the sense that we can access the defeating force they exert. Without taking sides on this debate here, any answer to that question plausibly is related to general questions about evidential support. If we think that a piece of evidence can support (and also potentially justify based on that support) some proposition p without the awareness of the believer, we are likely to think that this piece of evidence can also exert its defeating power independent of whether that defeating power is accessible to the believer (more on that below [§8]).
If we settle on the distinction between accessible and inaccessible mental state defeaters one way or another, one interesting question becomes whether there are defeaters which we possess and are not aware of but should be aware of. In other words, whether there is a second kind of normative defeater, which we may call normative mental state defeater or not:
Normative Mental State Defeater: Some evidence φ is a normative mental state defeater for S’s belief that p iff, (i) φ has enough defeating force to render S’s prima facie justified belief that p unjustified and (ii), φ is possessed, accessible and not accessed but should have been accessed.
I think Jennifer Lackey, on various occasions, has convincingly argued that there are cases involving what I call normative mental state defeaters.Footnote 17 I will discuss one such case and take a closer look at Lackey’s understanding of normative defeat in [§4.2.2].
In sum, this gives us two general types of defeaters, propositional and mental state defeaters, as well as various subtypes which inhabit different subparts of the entire evidence relevant to the epistemic situation ETOTAL. Similar to how we characterised different subtypes of the relevant evidence above [Fig. 1], we can plug in different kinds of defeaters into a tree diagram [Fig. 2]:
Figure 2. Defeaters.
Having categorised different types of defeaters, we can now investigate how defeaters exert their defeating force.
First, we need to consider the question of whether defeaters need to have some positive epistemic standing. Does φ, to have a defeating force towards Bp, need to be sufficient to support B¬p or at least to repudiate the support that S had for p in the first place? Here the traditional answer is that a defeater is a reason or piece of evidence φ such that given the initial evidence E and φ together are not sufficient to support p (while E alone would be) (Pollock Reference Pollock1986; Pollock and Cruz Reference Pollock and Cruz1999: 195; Graham and Lyons Reference Graham, Lyons, Brown and Simion2021: 40–41). Others have argued that mere beliefs without any positive epistemic standing can defeat otherwise justified beliefs (Plantinga Reference Plantinga2000: 364–5; Lackey Reference Lackey1999, Reference Lackey2008: 44–45; 60–63). For Lackey, independently of whether doxastic defeaters are ‘true, justifiedly believed, rationally believed’, what makes ‘an undefeated doxastic defeaters are epistemically problematic is that it is held in conjunction with another belief’ (Reference Lackey2005: 47). So, since contradicting unjustified beliefs can generate doxastic incoherence and if doxastic incoherence amounts to defeat, both justified beliefs and unjustified beliefs can serve as defeaters.Footnote 18
I am, following Graham and Lyons (Reference Graham, Lyons, Brown and Simion2021), sceptical of the idea that the doxastic incoherence generated by unjustified beliefs can repudiate the justificatory status of well-supported beliefs. Because I think that the correct response to this doxastic incoherence is to drop the unjustified belief, rather than both beliefs or merely the justified one (Graham and Lyons Reference Graham, Lyons, Brown and Simion2021: 49–50). Nevertheless, I acknowledge that this is an ongoing debate and that there are interesting arguments to be made in favour of both views. Accordingly, the following investigations rest on two debated assumptions. First, I will assume that defeaters need to have positive epistemic standing to be able to exert their defeating force, and second, that it is evidence, not beliefs (unless those beliefs amount to evidence), which exerts defeating force. These assumptions will not only enable me to maintain the above-proposed analogy between different types of evidence and different types of defeaters [§2] – [§3] but also link the notion of defeating force directly to the notion of evidential support and the evidentialist higher-order clause I will end up defending in [§8].
Following Pollock (Reference Pollock1986) I will assume that nothing can defeat a belief that cannot also provide justification. So, defeaters are either bodies of evidence strong enough to support justified beliefs or themselves justified beliefs (which, given my preferred understanding of evidence, also amount to evidence). As such, there are different ways defeaters can exert their defeating force. Both mental state and factive defeaters can be either rebutting, that is, provide evidence that the belief that p is false, or undercutting, that is, provide evidence that the belief that p is unreliably formed or sustained (Pollock Reference Pollock1986). In Pollock’s words, a rebutting defeater attacks the conclusion while an undercutting defeater ‘attacks the connection between the evidence and the conclusion, rather than attacking the conclusion itself’’ (Reference Pollock1986: 38). For example, reliable testimonial evidence that my colleague is currently in France is a rebutting defeater to my belief that I passed by her on campus today. In contrast, evidence about the unreliability of my facial recognition abilities, such as evidence that I have prosopagnosia or evidence about the unreliability of my short-term memory, is an undercutting defeater.Footnote 19
Moreover, a defeater may itself be defeated (Lehrer and Paxson Reference Lehrer and Paxson1969: 228–9). In such circumstances, we may speak of a defeated defeater who is defeated by a defeater-defeater. For example, the allegedly reliable testimonial evidence that defeated my belief that my colleague is currently in France might itself be defeated by hearing from multiple independent sources that the person whose testimony I am relying on is a notorious liar. Differentiating between defeated and undefeated defeaters is important since it is often argued that while justified beliefs are incompatible with defeaters, we only need to worry about defeaters who have not been defeated themselves, that is, undefeated defeaters. (I will question this assumption below [§4].)
In sum, there are various kinds of defeaters including propositional (normative and non-normative) and mental state (including doxastic, reflective, inaccessible, and normative) defeaters, which can defeat an otherwise justified belief in various ways (rebutting, undercutting, higher-order defeat), and which can itself be defeated. I will spend the next sections discussing conventional evidentialist [§4] and responsibilist [§5] strategies to account for the phenomenon of epistemic defeat. Based on these discussions, I will introduce a general taxonomy of defeater cases in [§6] which I will use to support my preferred understanding of defeat [§7] – [§8].
4. Evidentialism and defeat
In this section, I will summarise conventional evidentialist ways of handling defeat and list some well-known, as well as some novel shortcomings of these approaches. Let me start with some preliminaries about evidentialism as understood by Conee and Feldman (Reference Conee and Feldman2004).
The central evidentialist commitment in Conee and Feldman’s (Reference Conee and Feldman2004) framework is the following thesis:
Evidentialist Justification: A doxastic attitude D towards proposition p is epistemically justified for S at t if and only if having D towards p fits the evidence S has at t.
Sometimes the core commitment of evidentialism is also expressed as a supervenience thesis, which states that normative facts about the doxastic attitudes of the epistemic agent directly supervene on facts about their evidence. That is, any two epistemic agents possessing exactly the same evidence would be exactly alike concerning what they are justified in believing about any given issue (Kelly Reference Kelly2016: §1; Fratantonio Reference Fratantonio and Sylvanforthcoming: §1; Conee and Feldman Reference Conee and Feldman2004: 101).
However, many evidentialists take Evidentialist Justification as expressing only a necessary condition of epistemic justification rather than a sufficient condition, because even if the evidence possessed by an epistemic agent overall supports a given proposition, the way in which the respective attitude is formed can influence its normative status. So, if a belief is arrived at in an epistemically defective way, e.g., via wishful thinking, it is not fully justified even if it is supported by the possessed evidence. To this end, we need to distinguish between propositional and doxastic justification, where evidence alone only determines whether a belief is propositionally justified, for it to be doxastically justified, the belief also needs to be properly based on the evidence.Footnote 20 This gives us the following supervenience thesis:
Propositional Supervenience: The propositional justification of anyone’s doxastic attitude toward any proposition at any time strongly supervenes on the evidence that the person has at the time.
So, how does this Conee and Feldman-style evidentialism do with respect to some simple mental state defeater cases?
DEFEATER: S’s belief that p is prima facie justified. However, S is not aware that she possesses a reflective defeater φ for her belief.
DEFEATER-DEFEATER: S’s belief that p is prima facie justified. However, S is not aware she possesses a reflective defeater φ for her belief. Luckily, φ is itself defeated by another reflective defeater-defeater λ.
The Conee and Feldman-style evidentialist seems to have a convenient way to account for mental state defeater cases such as DEFEATER and DEFEATER-DEFEATER. They can appeal to the notion of propositional justification. If the entire evidence possessed by an epistemic agent needs to support or fit the respective belief for the belief to be propositionally justified the belief cannot be defeated by a mental state defeater. Since propositional justification is usually understood to be a necessary requirement for doxastic justification, S’s belief cannot be justified. So strictly speaking, for Conee and Feldman, there are no mental state defeaters since there is no (propositional) prima facie justification if the belief is not supported by the entire evidential base of S.
Since we are looking for a no-defeater clause that can be added to all kinds of understandings of prima facie justification, we need to transform this insight into an independent no-defeater clause:
Evidentialist No-Defeater Clause <Evidentialism>: S’s prima facie justified belief that p is undefeated iff S’s evidential base on balance supports p.
According to <Evidentialism> in DEFEATER-DEFEATER, S is justified in believing that p while in DEFEATER S’s belief that p lacks justification. Despite this being the seemingly desired result, I will demonstrate that handling defeater cases, via appeal to propositional justification is mistaken. In particular, I will discuss two kinds of propositional defeater cases [§4.1] and two kinds of mental state defeater cases [§4.2] on which <Evidentialism> fails to deliver the right verdict.
4.1. Evidentialism and propositional defeaters
My first thesis is that, since <Evidentialism> is only concerned with defeaters within the evidential base, it cannot handle cases of propositional or normative defeat.
PROPOSITIONAL DEFEATER: S’s belief that p is prima facie justified. However, S is not aware that there is an unpossessed but available defeater φ for her belief that p.
NORMATIVE DEFEATER: S’s belief that p is prima facie justified. However, S is not aware that there is an unpossessed but available defeater φ for her belief that p that S should have possessed.
Depending on whether S should have possessed φ or not we have a template to construct normative propositional or merely propositional defeater cases. I will, for now, focus on normative defeaters. However, I will revisit this distinction at the end of the section. (Furthermore, the account I’ll offer in [§8] can make sense of normative and non-normative propositional defeater cases).
The unique threat normative defeaters have for evidentialism has been pointed out by many, including Kornblith (Reference Kornblith1983), Baehr (Reference Baehr2009), Cloos (Reference Cloos2015), as well as Goldberg (Reference Goldberg2017: 2891–3; Reference Goldberg2018: ch.6). If we are merely concerned with the overall evidence an epistemic agent possesses at a time, we can never consider the evidence (defeating or not) that is not possessed by S, but should be possessed by S. One illustrative example that has the structure of NORMATIVE DEFEATER is found in Kornblith (Reference Kornblith1983):
PERSISTENT PETE: Pete a young and stubborn physicist presents his newest pet hypothesis at a conference eager to hear the praise of his colleagues. Due to a personality disorder, Pete pays no attention to reasonable critics and strategically ignores important counterevidence. As a result, even devastating criticism fails to impact his beliefs not because he fails to take it into account but because he has not even heard it.
For Kornblith this example illustrates that Pete’s “belief is unjustified, after his colleague presents his objection, and it is unjustified because of his culpable ignorance’ (Kornblith Reference Kornblith1983: 36). As such, Kornblith offered the case as a direct challenge to evidentialism, since if his interpretation is correct there are non-evidential factors which determine whether Pete is justified or not.Footnote 21 While evidentialists have tried to dismiss the intuitive judgement, we have towards cases such as PERSISTENT PETE, none of their strategies seems particularly convincing. For example, Conee and Feldmann (Reference Conee and Feldman2004: ch.7) have argued that cases of normative defeat do not undermine epistemic justification but usually demark other shortcomings such as moral or professional failures. However, many disagree with this judgement (Goldberg Reference Goldberg2016: 450; Lackey Reference Lackey2016: 374–5, Graham and Lyons Reference Graham, Lyons, Brown and Simion2021). After all, having these prudential or professional failures has devastating epistemic effects on Pete, not only concerning the reliability and truth-conduciveness of his belief-forming mechanisms but also concerning the evidence he possesses. In other words, there are good reasons to think that these allegedly professional failures are also epistemic failures (more on that below).
At this point, a plausible suggestion is to expand the evidence we are concerned with to include unpossessed evidence. That is, we might require that the total relevant evidence, possessed and unpossessed, on balance supports p.
Extended Evidentialist No-Defeater Clause <Extended Evidentialism>: S’s prima facie justified belief that p is undefeated iff the total evidence on balance supports p.
While <Extended Evidentialism> accounts for cases such as PERSISTENT PETE and other cases of normative defeat, it already demarks a significant departure from Conee and Feldman’s version of internalist evidentialism, since it denies the central supervenience claims. Furthermore, it cannot account for the following type of normative defeater cases:
LUCKY NORMATIVE DEFEATER-DEFEATER: S’s belief that p is prima facie justified. However, S is not aware that there is an unpossessed but available defeater φ for her belief that p that S should have possessed. Luckily for S, there is another normative defeater λ which defeats φ.
To pump some intuitions, it might help to put some flesh on the bones:
SORROWLESS SARAH: Sarah forms a prima facie justified belief that p. However, the belief is based on some information E drawn from the daily tabloid, unknowingly to Sarah a very unreliable source. This fact constitutes a normative defeater φ for the belief that p, since Sarah could have easily and should have learned about φ. Instead of checking the quality of her sources Sarah unreflectively formed her belief based on E. Yet, while generally unreliable the daily tabloid is reliable in this particular instance since the one columnist Sarah based her belief on is exceptionally reliable. This fact serves as a defeater-defeater λ for φ.
Sarah’s prima facie justified belief appears to be defective because she was just lucky that the defeater that she should have possessed is itself defeated. She could have easily learned about φ which, if rational, would have led her to abandon the belief. In not possessing φ Sarah behaved epistemically irresponsible just as Pete behaved epistemically irresponsibly when ignoring the counterevidence presented by his colleagues.
If that’s the right verdict, we can easily see that extending <Evidentialism> to <Extended Evidentialism> does not solve the problem of normative defeat. Instead, normative defeater cases seem to suggest that we require epistemic agents to be responsible in the right kind of way. This is also illustrated by the following type of propositional defeater cases:
UNLUCKY PROPOSITIONAL DEFEATERS: S’s belief that p is prima facie justified. However, while being generally a very responsible investigator S is not aware that there is an unpossessed and available defeater φ for the belief that p.
Here is such a case:
METICULOUS MICHAEL: Michael’s belief that the tip of his tongue is mainly responsible for tasting sweet is prima facie justified. Michael has good evidence that this is true. He has once learned it in school and even remembers (apparently) confirming it himself as a kid. Furthermore, he even double-checks his belief by looking it up in his old biology book. However, while being generally a very responsible investigator Michael could have easily found out (e.g., via a quick Google search) that the tongue-tasting map is a common scientific misconception that has repeatedly been proven wrong. This fact serves as a propositional defeater φ for Michael’s belief.
Can Michael’s belief that p be justified despite there being a propositional defeater φ? It seems like it makes a difference if Michael acts meticulously or not. After all, he double-checked his belief and in general has good evidence for it being true. At least, there seems to be a difference to cases such as PERSISTENT PETE. Pete willfully ignored counterevidence and formed his belief partly based on the desire to prove his pet hypothesis.
If we judge Pete’s and Michael’s beliefs to have a different epistemic status, there are at least two ways to explain the different results. First, we might think that the way Pete and Michael form their beliefs makes those beliefs epistemically responsible or irresponsible respectively and hence justified or unjustified. On the other hand, we might think that the different judgements suggest that PERSISTENT PETE is a normative defeater case, where Pete should have possessed φ, while METICULOUS MICHAEL is merely a propositional defeater case, and hence Michael had no epistemic obligation to possess φ. Either way, the analysis goes through the notion of epistemic responsibility, which suggests a responsibilist treatment of defeat. I will explore this responsibilist treatment of defeaters further in [§5]. However, before plunging ahead, let me discuss another set of defeater cases evidentialist approaches have trouble dealing with [§4.2].
4.2. Evidentialism and mental state defeaters
While <Evidentialism> could handle regular mental state defeater cases, such as DEFEATER or DEFEATER-DEFEATER, normative and propositional defeater cases illustrated that <Evidentialism> is too narrow. In trying to overcome this problem by extending <Evidentialism> to <Extended Evidentialism>, we made it too general, since it declared sorrowless Sarah’s belief to be justified despite her epistemically irresponsible behaviour and meticulous Michael’s belief to be unjustified despite his epistemically responsible behaviour. We can find a similar pattern when we take a closer look at mental state defeater cases.
In particular, if we use propositional justification to declare all prima facie justified beliefs in all mental state defeater cases to be unjustified and in all defeater-defeater cases to be justified, we get the wrong verdict with respect to two cases: inaccessible defeater cases, in which the epistemic agent is responsible [§4.2.1] and accessible defeater-defeater cases, in which a defeated defeater is irresponsibly ignored [§4.2.2]. Let me discuss those cases subsequently.
4.2.1. Inaccessible defeaters
If we solely care about propositional justification, we rule out justification in any case in which we stipulate that there is an (undefeated) defeater φ within EBASE. This is the case, irrespective of how strong the initial justification for Bp is (as long as φ has enough defeating force), regardless of whether the defeating evidence is reflectively accessible or entirely inaccessible, and independently of how responsible the epistemic agent is. Here is a template for such cases:
UNLUCKY MENTAL DEFEATERS: S’s belief that p is prima facie justified. However, while being generally a very responsible investigator S is not aware that she possesses a non-doxastic inaccessible or hardly accessible defeater φ for the belief that p.
We can utilise this template to get the following case which bears some similarities to METICULOUS MICHAEL [§6.4.1]:
CAREFUL CAROLINE: Caroline’s belief that the tip of her tongue is mainly responsible for tasting sweet is prima facie justified. Caroline has good evidence that this is true. She has once learned it in school and even remembers (apparently) confirming it herself as a kid. Furthermore, she even double-checks her belief by looking it up in her old biology book. However, while generally being a very responsible investigator, Caroline forgot that her university professor once told her that the tongue-tasting map is a common scientific misconception that has repeatedly been proven wrong. This currently inaccessible memory serves as a non-doxastic defeater φ for Caroline’s belief.
CAREFUL CAROLINE aims to illustrate that it depends on the exact details of mental state defeater cases if we should consider the apparently defeated beliefs to be justified or not. It seems overly demanding to require that epistemic agents are always completely sensitive to the entire body of evidence they possess, independently of how accessible the defeater is. After all, as justification is fallible concerning the truth of the formed beliefs, it is also plausibly fallible with respect to the overall evidential support.
So, depending on the exact nature of the evidential base, <Evidentialism> may lead to an absurdly restrictive notion of justification. Take, for example, the so-called inclusive view of evidence discussed by Conee and Feldman (Reference Conee and Feldman2004: 228). On this view, the evidence possessed by an epistemic agent is just the sum of all their non-factive mental states. That means the evidential base includes wrong beliefs and unconscious or in-principle inaccessible mental states. All of these mental states could serve as defeaters. In other words, even if S is overly responsible, attentive and aware of most of the evidence they possess the prima facie justified belief that p could ultimately be unjustified because they happen to possess an inaccessible or hardly accessible defeater for the respective belief.
Confronted with these cases, the evidentialist might be tempted to restrict the notion of the evidential base in a way that rules out inaccessible or hardly accessible evidence. However, then the problem posed by normative defeaters becomes all the more important. In restricting the evidential base, we would push the defeater out of it into the unpossessed but available or required evidence and thereby transform cases such as CAREFUL CAROLINE into normative defeater cases such as PERSISTENT PETE.
4.2.2. Normative mental state defeaters
Furthermore, both <Evidentialism> and <Extended Evidentialism> also give us the wrong verdict for some normative mental state defeater-defeater cases. To see this, let us start with this case inspired by Lackey (Reference Lackey2008: 63; see also Reference Lackey1999: 487):
IGNORANT ALICE: Alice is told by her optometrist that her vision is nearly completely unreliable, yet she refuses to accept his diagnosis, without having any rational basis for doing so. Afterwards, as she is walking out of the doctor’s office, she sees a car accident. Based on that Alice forms the corresponding true belief that there was such an accident. Intuitively, however, her belief is defeated by the doctor’s diagnosis even if she refuses to accept it.
Let us compare this case to some cases already discussed. On the face of it, IGNORANT ALICE looks like a regular mental state defeater case rather than a normative defeater case such as PERSISTENT PETE. That is, ignorant Alice, in contrast to persistent Pete, possesses but ignores a defeater φ. Following Lackey, we could, nonetheless, call these kinds of defeater cases normative since Alice should form the corresponding belief. More precisely, Lackey thinks that a ‘normative defeater […] function by virtue of being doubts or beliefs that S should have (whether or not S does have them) given the presence of certain available evidence’ and further, that they exert their defeating force because ‘certain kinds of doubts and beliefs—either that one has or should have—contribute epistemically unacceptable irrationality to doxastic systems and, accordingly, knowledge (justification/warrant) can be defeated or undermined by their presence’ (Reference Lackey2008: 45; emphasis in the original). So, in this particular case, for Lackey, Alice should form the belief that her visual perception is unreliable because she possesses good evidence provided by the optometrist that would support this belief, whereas this belief would serve as a defeater, once formed.
While I agree with Lackey that IGNORANT ALICE constitutes a normative defeater case, I disagree about the way the testimony Alice received from her optometrist defeats Alice’s belief about the accident. It is not the belief that Alice should have formed but the testimonial evidence that serves as a defeater φ directly. This falls out of my view on how to understand defeating force [§3]. E (the witnessing of the car accident) together with φ (the testimony of the optometrist) do not together support p (that there was a car accident) and hence φ defeats Alice’s belief that p.Footnote 22 This leaves us with the observation that we seem to have special normative obligations towards not only some evidence we should possess (as in normative defeater cases) but also some evidence we already possess but should access, which I call normative mental state defeater cases.
How does <Evidentialism> do with respect to normative mental state defeater cases, such as IGNORANT ALICE? At first, it seems like evidentialists can make perfect sense of them, since, after all, Alice’s belief is not propositionally justified. Nonetheless, we can use the idea of having normative obligations to access some bits of our evidential base to construct the following type of defeater-defeater case which poses a problem for <Evidentialism>:
LUCKY MENTAL DEFEATER-DEFEATER: S’s belief that p is prima facie justified. However, due to being ignorant and negligent, S is not aware that she possesses an accessible defeater φ which she should have accessed for the belief that p. Luckily, S also possesses an inaccessible or hardly accessible defeater-defeater λ for φ.
We can again add some flesh to the bones:
CARELESS CARL: Carl’s belief that he saw his colleague Hao on campus today is prima facie justified. Carl has a vivid memory of her passing by the café while he was drinking his after-lunch coffee. Carl when asked if Hao is in town responds that she is without further reflection. However, another colleague told Carl last week that Hao will fly out on holiday tomorrow, something that Carl could recall easily if he would just reflect on his belief. This would make Carl doubt his memory and likely lead him to judge that he must have mistaken Hao for her twin sister. Accordingly, this memory counts as a defeater φ for Carl’s belief that Hao is in town. Luckily for Carl, he also possesses a defeater-defeater λ for φ; namely that Hao told him that she did not plan any holiday this year and that she will definitely stay in town over the summer to work on her newest book. However, in contrast to φ, Carl cannot easily recall λ and hence would only remember it after a long period of reflection.
CARELESS CARL illustrates that not only in defeater cases but also in defeater-defeater cases the omitted details matter. Carl’s belief, while undefeated and prima facie justified, seems epistemically defective for various reasons. First, Carl’s belief is unstable. He could easily be aware of φ which, if rational, would lead him to dismiss the belief. So, it seems wrong to judge Carl’s belief to be justified because of the luck involved. Carl, while having prima facie justified beliefs clearly forms his beliefs improperly. It is a mere coincidence that Carl’s otherwise defeated belief ends up being undefeated. So, it seems like we have all the right to blame Carl for having this belief and generally forming beliefs in that way. And if this blameworthiness goes hand in hand with our epistemic evaluations, we need to say that Carl’s belief is unjustified.
In sum, conventional evidentialist strategies fail to give us the right verdict in at least two different kinds of defeater cases. First, they are insensitive to propositional and normative defeaters, and second, they are too hasty in declaring all mental state defeater cases to be unjustified and all mental state defeater-defeater cases to be justified. While extending the pool of considered evidence (such as in <Extended Evidentialism>) helps with some propositional and normative defeater cases it does not help with the latter types of cases discussed.
5. Responsibilism: Two desiderata
As demonstrated above, conventional evidentialism has problems with a plethora of defeater cases; namely, normative defeater cases (PERSISTENT PETE), normative defeater-defeater cases (SORROWLESS SARAH), lucky propositional defeater cases (METICULOUS MICHAEL), inaccessible mental state defeater cases in which the epistemic agent is otherwise very responsible (CAREFUL CAROLINE), and defeater-defeater cases in which the agent behaves in an epistemically defective way (CARELESS CARL). On the face of it, these misjudgements of evidentialist strategies stem from a common source: they do not take into account whether the epistemic agents behave epistemically responsible or not.
Accordingly, one way to account for these cases is to introduce a no-defeater clause that obliges epistemic agents to be epistemically responsible in the right kind of way. On similar grounds, many have argued that we have epistemic normative obligations (Lackey Reference Lackey2016; Reference Lackey2021; Silva Reference Silva2019), requirements to be sensitive to the evidence we should have had (Goldberg Reference Goldberg2016; Cloos Reference Cloos2015) or that we are required to know everything we should have known due to professional or prudential duties (Goldberg Reference Goldberg2017). Based on these insights, Goldberg proposes an account of epistemic responsibility that is grounded in the epistemic expectations that epistemic agents are entitled to have towards each other (Goldberg Reference Goldberg2018; ch.6). While I am sympathetic to this so-called entitlement model of epistemic responsibility, I will argue in [§8], that is entitlement is best understood as amounting to an entitlement to expect others to possess and utilize higher-order evidence that bears on their epistemic situation.
For now, however, we can lump these strategies together to specify a responsibilist treatment of defeat as follows:
Responsibilist No-Defeater Clause <Responsibilism>: S’s prima facie justified belief that p is undefeated iff S is epistemically responsible in believing that p.
<Responsibilism> looks promising, and indeed, if understood correctly, it gives us the right verdict with respect to all of the cases discussed so far [§6]. However, as I will demonstrate in [§7] and [§8], there are more and less plausible ways to understand epistemic responsibility. For now, I will just, based on the discussions above, identify two desiderata any account of epistemic responsibility needs to fulfil:
Desideratum 1: Any responsible epistemic agent forms and maintains doxastic attitudes in a way that is sensitive to the evidence they accessed and should have accessed.
Desideratum 2: Any responsible epistemic agent forms and maintains doxastic attitudes in a way that is sensitive to the evidence they should have possessed.
These desiderata, while falling short of giving us a full-blown understanding of epistemic responsibility, will be helpful when comparing different kinds of no-defeater clauses by checking their verdicts concerning the full taxonomy of defeater cases [§6].
Before moving on, however, let me briefly introduce a hybrid no-defeater clause that is neither exclusively evidentialist nor exclusively responsibilist.
Responsibilist-Evidentialist No-Defeater Clause <Responsible Evidentialism>: S’s prima facie justified belief that p is undefeated iff S’s evidential base on balance supports p and S is epistemically responsible in believing that p.
Both Baehr (Reference Baehr2009) and Cloos (Reference Cloos2015) point out that there is no conflict, in principle, with combining responsibilist notions of defeat with evidentialist notions of justification.Footnote 23 However, their accounts are full-blown accounts of justification rather than merely no-defeater clauses that are used to supplement understandings of prima facie justification. In contrast, <Responsible Evidentialism> can be added to all kinds of understandings of prima facie justification including evidentialist, responsibilist, or reliabilist accounts.
In sum, in [§3] – [§5], I have identified four responsibilist and/or evidentialist no-defeater clauses and discussed various defeater cases. The next section uses these insights to give us a general taxonomy of defeater cases against which any no-defeater clause can be tested.
6. Responsibility, balance and location: A taxonomy
Most of the proposed no-defeater clauses make different judgements about the justificatory status of beliefs depending on the location of the defeater. For example, for <Evidentialism> and <Responsible Evidentialism> whether the belief that p is justified (in part) depends on whether the defeater is possessed or unpossessed (within or outside of EBASE). <Extended Evidentialism> in contrast declares the belief to be defeated as long as there is a defeater within the total evidence ETOTAL. Let us call the parameter that demarcates whether the defeater is possessed or unpossessed the location of the defeater.
In addition to the location, most evidentialist accounts care about the propositional justification, that is, the overall balance of the evidence. If there is a defeater φ (within EBASE), the belief is unjustified; however, if there is a defeater-defeater λ that undermines φ the belief’s justification is restored again. That is, the overall evidence can be in favour of p because there is no defeater, or because there are only defeated defeaters. Let us call this parameter the evidential balance.
<Responsibilism> declares beliefs to be un/justified regardless of the overall balance and location of the defeater(s), while <Responsible Evidentialism> declares Careless Carl’s belief to be unjustified and Careful Caroline’s to be only justified if the defeater is not part of her evidential base. In contrast, <Evidentialism> and <Extended Evidentialism> do not take the careless or careful nature of the epistemic agent into account at all. Let us call this parameter the responsibility of the epistemic agent.
Having identified the parameters, responsibility, balance, and location presents us with eight possible cases, which, in turn, gives us a general taxonomy of defeater cases. Here is a template, we can use to change the parameters accordingly to see which account gives us which result:
Defeater Template: S’s belief that p is prima facie justified. There is a [balance: defeated/undefeated] and [location: possessed/unpossessed] defeater φ for S’s belief that p. S is [responsibility: careless/careful] in forming and sustaining the belief that p.
In [Fig. 3] I plugged in the judgements made by <Evidentialism>, <Responsible Evidentialism>, <Extended Evidentialism>, and <Responsibilism> with respect to the different possibilities provided by Defeater Template, whereby ‘N’ stands for defeated or unjustified and ‘Y’ stands for justified or undefeated:
Figure 3. Taxonomy.
By plugging in the respective parameters into the Defeater Template, we can see that only <Responsibilism> matches our intuitive judgements: careless Carl’s belief (careless; defeated; possessed) is not justified, such as persistent Pete’s and sorrowless Sarah’s (careless; defeated; unpossessed), while Careful Caroline’s belief (careful; defeated; possessed) and meticulous Michael’s belief (careful; defeated; unpossessed) remain justified [Fig. 3].
While this sounds like a full-blown responsibilist conclusion, the understanding of epistemic responsibility sketched in [§5] was quite schematic and hence is in need of further clarification. I will spend the next section [§7] adding further details to the proposed understanding of epistemic responsibility, while I will argue in [§8] that the best way to understand epistemic responsibility is as a higher-order evidentialist requirement.
7. Responsibilism revisited
In [§5] I have identified two desiderata underlying any responsibilist treatment of defeat that is able to give us the correct verdict with respect to the full range of defeater cases [§6].
While these two desiderata were helpful in specifying which shape any no-defeater clause should have, we lack a full account of epistemic responsibility. As a first step, we could refer to the above-sketched understanding of epistemic responsibility as being grounded in socio-epistemic expectations (Goldberg Reference Goldberg2016; Reference Goldberg2017; Reference Goldberg2018). This would allow us to make things more precise:
Epistemic Responsibility: S is epistemically responsible if S is sensitive to (i) the evidence S possesses, including those which S has accessed and should access according to basic normative expectations, and sensitive to (ii) the evidence S should possess according to the epistemic expectations we are entitled to have towards S.
However, even if we accept Golberg’s understanding of epistemic responsibility as being grounded in socio-epistemic expectations, a big question mark still hangs over the proposed understanding; namely, we still need to say what exactly it means for an epistemic agent to be sensitive towards defeaters in the right way.
Here, I think, as I have argued elsewhere (Graf Reference Graf2024: 176–85), extant understandings of epistemic responsibility, such as the one proposed by Goldberg, give us the wrong answer to this question. In short, Goldberg thinks that epistemic agents are strictly liable to the evidence they possess as well as the evidence they should have possessed (Goldberg Reference Goldberg2018: 215), which gives us the wrong result with respect to some of the above-discussed cases. But instead of engaging in comparative analysis here, I want to put forward a positive proposal [§8].
8. A higher-order evidence clause
To recapitulate: we want a no-defeater clause that makes epistemic agents responsible, that is, sensitive to evidence, possessed or unpossessed, they should take into account, in a way that makes room for lucky and unlucky cases. In so doing, we want epistemically responsible agents to be less likely to ignore defeaters but rule in cases in which they are responsible but miss some defeaters due to no fault of their own. Furthermore, we want to rule out cases in which epistemic agents are irresponsible in forming some beliefs but are lucky that those beliefs are not defeated.
One promising way to retain these intuitive judgements about epistemic responsibility is to equate being epistemically responsible with possessing and utilising higher-order evidence about one’s epistemic situation. That is, instead of taking epistemic agents to be liable for all of the evidence they have and should have had, we expect that epistemic agents have and utilise higher-order evidence about the entire evidence relevant to their epistemic situation ETOTAL. So, we are not only entitled to expect people to not ignore their higher-order evidence about unpossessed and unaccessed evidence, but we are entitled to expect epistemic agents to have such evidence in the first place. This enables us to distinguish between lucky and irresponsible formed beliefs, such as careless Carl’s beliefs, as well as unlucky and responsible formed beliefs, such as careful Caroline’s. We judge Carl to be irresponsible because he does not have and utilise higher-order evidence while we judge Caroline responsible because she does.
This suggests that being sensitive and hence epistemically responsible really just is possessing and utilizing higher-order evidence in the right kind of way. Let me further illustrate this idea with another example. For instance, we trust a doctor who makes a diagnosis based on some blood test result because of the doctor’s expertise in reading and interpreting blood test results. This means that we are entitled to expect the doctor to have higher-order evidence supporting the general reliability of blood test results and his ability to interpret them. That is, we expect doctors to possess general information about blood tests via their medical education and specific information about their track record of making diagnoses based on blood tests. After all, if the doctor does not possess or use any such higher-order evidence in making their diagnosis we are entitled to blame them for that. So, for example, if the doctor has not had enough practice in making a diagnosis on blood test results, we expect them to indicate that. That is independent of whether the doctor’s failure to use any such higher-order evidence leads him to misjudge the blood test results or not.
This indicates that we should not understand epistemic responsibility, while grounded in social expectations, via liability.Footnote 24 The doctor is not liable for not having and utilising evidence he should have in the case this leads to a misjudgement of the situation but is generally required to have higher-order evidence about her situation. Let us make this more precise. Let us stipulate that Dr. S looked at the blood test results E and concluded that p. Furthermore, Dr. S’s belief is also based on a bunch of background evidence EH about the general reliability of blood tests and her competence with respect to interpreting these test results. That is, E, as well as EH played a non-deviant causal role in the belief formation of S’s belief and/or S would provide or has the disposition to provide E and EH as reasons for p if asked (depending on the underlying understanding of epistemic basing).
In sum, this gives us the following picture of epistemic responsibility. First (i) running the risk of missing relevant evidence by not acquiring higher-order evidence about one’s epistemic situation suffices for epistemic irresponsibility. Second, (ii) not running the risk but nonetheless missing evidence does not suffice for epistemic irresponsibility. Where responsibility is a necessary condition for having an undefeated and hence justified belief. This is the understanding of epistemic responsibility that underlies the higher-order evidentialist no-defeater clause announced above [§1]:
-
Higher-Order Evidence Clause <HOE>:
-
S belief that p is undefeated iff:
-
(a) S possesses some higher-order evidence EH, which is sufficient to support q, the proposition that the total relevant evidence ETOTAL on balance supports p and
-
(b) S’s belief that p is properly based on EH.
-
Let me now take a closer look at both subclauses of <HOE> and illustrate how they can be interpreted differently depending on different underlying understandings of evidence, evidential support, and epistemic basing.Footnote 25
-
(a) states that epistemic agents need to have higher-order evidence EH that is sufficient to support the proposition q, the total relevant evidence on balance supports p. So, one promising way to understand clause (a) is that epistemic agents need to have higher-order evidence in the form of a mental state (with the relevant propositional content) that supports q, which means sufficiently increases the probability of q, that ETOTAL supports p.
-
(b) encompasses the idea that merely stipulating that S possesses some second-order evidence EH is not sufficient since this does not guarantee that EH is involved in the formation of the belief that p. Without (b), EH could not bear on the epistemic standing of Bp. So, as with beliefs formed via first-order evidence, we need a basing requirement. To illustrate this, we might think about careless Carl and careful Caroline again. If careful Caroline is responsible by having and utilising EH it does not matter whether there is a defeater (possessed or unpossessed) for her prima facie justified belief. She is responsible because she tries to ensure that there are no defeaters. And, she did so by utilising sufficient higher-order evidence in her reasoning. In contrast, Carl, who forms beliefs on a whim, is irresponsible, even if he does coincidentally possess some higher-order evidence EH. Subclause (b) encapsulates this intuition. It ensures that EH is not merely possessed but actually used. That is, based on different understandings of epistemic basing, clause (b) could be understood in multiple ways, it could, e.g., mean that the belief that p is (non-deviantly) caused by EH.
Note that given the above-outlined understandings of evidence, evidential support, and epistemic basing, accepting <HOE> as a necessary requirement on epistemic justification does not lead to any kind of over-intellectualization.Footnote 26 In other words, <HOE> is not very cognitively demanding, despite its higher-order nature. All it requires is that epistemic agents need to be capable of having two mental states with the relevant propositional content E and EH that are sufficient to support (which, e.g. means increase the probability of) two different propositions and that E and EH non-deviantly cause S to believe that p. None of this needs to be happening consciously, S could just be acquainted with a belief-formation procedure that utilizes higher-order and first-order evidence in the right way. This can happen without S being aware of E or EH as well as the evidential support that E and EH grant to the respective propositions p and q and without E and EH and their support being accessible to S.Footnote 27 This means that, in principle, not only adult human beings and other relatively cognitively sophisticated epistemic agents can form justified beliefs but also children and other (allegedly less sophisticated) non-human animals.
With these clarifications about (a) and (b), it is easy to demonstrate that <HOE> gives us the right verdict with respect to the entire taxonomy of defeater cases [§6]. Defeated beliefs can be justified independent of the overall evidential balance and the location of the defeater if the belief is partly based on higher-order evidence that sufficiently bears on the overall evidential situation. So, careless Carl and sorrowless Sarah have unjustified yet ‘undefeated’ beliefs, while careful Caroline and meticulous Michael have justified, yet ‘defeated’ beliefs. This is the case because one lacks while the other possesses and uses higher-order evidence that the total relevant evidence ETOTAL supports the formed doxastic attitude. That the higher-order evidence is actually (properly) used is important since it suggests that the agent in consideration is in fact responsible and not only possesses sufficient evidence.
We can now take a step back and reflect on the initial idea of the defeasibility of justified beliefs that motivated our search for a no-defeater clause more generally and our analysis of epistemic responsibility in specific.
9. Concluding thoughts
Our search for a no-defeater clause was motivated by the doctrine of Defeatism [§1], which is the idea that doxastic attitudes can have the status of being prima facie justified (by having some justification conferring property) while being defeated. Defeatism is motivated by a plethora of defeater cases involving various types of defeaters. While I have demonstrated that responsibilist ideas motivate a specific understanding of defeat that can deal with the entire taxonomy of cases discussed [§6], I have also argued that the best way to spell out the underlying desiderata of responsibilism is as higher-order evidentialist requirement [§8]. This is good news for those who prefer an evidentialist approach to justification in general. Especially, since many have argued that evidentialism cannot make sense of many wide-shared intuitions with respect to defeater cases. Nonetheless, <HOE> is different from both extant evidentialist as well as responsibilist treatments of defeat.
First, <HOE> is different from the other evidentialist clauses discussed, since it is like responsibilist requirements a positive rather than a negative no-defeater clause. As such, like other positive no-defeater clauses <HOE>, while motivated by defeater cases, does not necessarily commit oneself to the doctrine of Defeatism. After all, accepting <HOE> we do not need to distinguish between prima facie justified beliefs which can be defeated and ultima facie justified beliefs. Instead, higher-order evidential requirements could be understood as necessary conditions for justification and once a belief fulfils whatever the first-order conditions are plus the specified higher-order requirements the belief is justified, independently of whether there is a defeater or not (within or outside of EBASE). This is in stark contrast to extant evidentialist understandings of defeat. These understandings rely on the notion of prima facie justification, which they spell out (partly) in terms of propositional justification [§4]. <HOE>, in contrast, is compatible with rejecting both of these ideas: we neither need to distinguish between prima facie and ultima facie justification nor do we need to spell out prima facie justification in terms of propositional justification.
Second, <HOE> is importantly different from extant responsibilist no-defeater clauses. If we accept <HOE> as a general solution to defeater cases, we are not committed to the idea that there are epistemic obligations that require us to possess certain bodies of evidence (as argued by many responsibilists). Given <HOE>, there is no evidence we should – epistemically speaking – have possessed or have accessed (except the higher-order evidence EH that is required to be responsible).Footnote 28 In that sense, most of the distinctions drawn in [§2] and [§3] are redundant. The only set of evidence we need to define to analyse the phenomenon of epistemic defeat is ETOTAL, the entire evidence that is relevant to an epistemic situation, to judge whether some evidence EH is sufficient to support that ETOTAL is sufficient to support p.
Acknowledgements
This article has benefited considerably from discussions, comments, and feedback from Robbie Williams, Sandy Goldberg, Jessica Brown, Andrew Peet, and Joshua Habgood-Coote, as well as anonymous comments from an associate editor and a referee from Episteme. Some substantial revisions to this article were made while I was a short term visiting fellow at the Institute Vienna Circle. Accordingly, I am also grateful to Georg Schiemer and other members of the institute for their support.
Funding statement
The project leading to this paper has received funding from the European Research Council (ERC) under European Union’s Horizon 2020 research and innovation programme (grant agreement no. 818633).
Competing interests
The author has no competing interests to declare that are relevant to the content of this article.
Open access
