Moral Disagreement and the Significance of Higher-Order Evidence

the Significance of

Higher-Order Evidence

Marco Tiozzo

ISBN 978-91-7346-992-0 ISSN 0283-2380

Available online at: http://hdl.handle.net/2077/57974

Academic thesis in Practical Philosophy, at the Department of Philosophy, Linguistics and Theory of Science

ACTA UNIVERSITATIS GOTHOBURGENSIS

Box 222, SE-405 30 Göteborg, Sweden acta@ub.gu.se

Cover design by Marco Tiozzo.

Printed by Brand Factory, Kållered 2018

Title: Moral Disagreement and the Significance of Higher-Order Evidence

Author: Marco Tiozzo Language: English

ISBN: 978-91-7346-991-3 (printed) ISBN: 978-91-7346-992-0 (pdf) ISSN: 0283-2380

Keywords: moral skepticism, disagreement, higher-order evidence, epistemic

rationality, peer disagreement, faultless disagreement, moral intransigence

Recent years have seen an increasing interest in the philosophy of disagreement, especially in epistemology where there is an intense debate over the epistemic significance of disagreement and higher-order evidence more generally.

Considerations about disagreement also play an important role in metaethics – most prominently in various arguments that purport to establish moral skepticism.

This thesis presents five papers that address

moral disagreement and higher-order evidence. The

first two papers examine the significance of higher-

order evidence and do not directly discuss moral

disagreement but are still necessary in order to

develop the epistemological framework that is used

in the dissertation. The first paper considers

different explanations of higher-order defeat. It is

argued that higher-order defeat is contingent on

whether the subject in question comes to believe

that her belief about the relevant matter fails to be

rational. The Second paper suggests that

views about epistemic rationality.

The other three papers, consider specific arguments from disagreement: the Argument from Peer Disagreement; the Argument from Faultless Disagreement; and the Argument from Dogmatism.

The third paper argues that the Argument from Peer Disagreement fails to make a case for widespread moral skepticism; mainly because higher-order evidence only contingently leads to defeat. The fourth paper examines a recent attempt to epistemically account for faultless moral disagreement without giving up on moral realism.

The paper argues that this attempt to accommodate

faultless disagreement is unsatisfactory. The fifth

paper develops a new argument against

cognitivism: the Argument from Dogmatism. The

argument holds that the conceivability of moral

dogmatists, i.e., agents who stubbornly stick to

their moral judgments in the face of putative

counterevidence, gives us reason to think that

moral judgments are not evidence-sensitive in the

way beliefs are.

I. Higher-Order Defeat: An Explanatory Problem

Marco Tiozzo Submitted 2018

II. The Level-Splitting View and the Non- Akrasia Constraint

Marco Tiozzo

Forthcoming in Philosophia.

doi: 10.1007/s11406-018-0014-6

III. Moral Disagreement and the Limits of

Higher-Order Defeat

IV. Matters of Ambiguity: Faultless

Disagreement, Relativism and Realism John Eriksson and Marco Tiozzo

Originally published in Philosophical Studies 173 (6):1517-1536, 2016.

doi: 10.1007/s11098-015-0561-7

V. The Moral Dogmatist: A Challenge to Cognitivism

John Eriksson and Marco Tiozzo Manuscript

Previously published papers are reprinted with

permission.

To paraphrase Otto Neurath: writing a dissertation is in many ways like reconstructing a ship on the open sea and never being able to start afresh from the bottom. It has been a long and hard ride but I am slowly starting to see the light at the end of the tunnel. I have spent a great deal of time thinking and writing about moral disagreement. The dissertation did not leave me with much time to do anything else. I miss the peaceful enjoyment of reading fiction and look forward to having more time for earthly pleasures. But like it says in the Bible, “there is a time for everything”. I think that the first thing I will do when I have handed in this thesis for printing is to go a day without doing anything, maybe go for a beer or two.

I am grateful to many people. First of all, I would like to thank my supervisors Ragnar Francén and Caj Strandberg for their invaluable support during the entire process of writing this dissertation.

Thank you Caj for succinct and often crucial

comments on parts of the manuscript. Thank you

Ragnar for all the patient readings and sharp

comments on my writings. Without your relentless

work, especially during this last year, I am not sure

that the dissertation would have seen the light of

day.

have offered stimulating discussion on various topics (not only philosophical) over these past years. I hope that our philosophical disagreements have been as productive for you as they have been for me. I am certain that they have helped me make progress as a philosopher.

I would also like to thank the numerous people who have commented on my work during these years. The participants to the seminar in practical philosophy, but also participants to various workshops and conferences in Cologne, Copenhagen, Linköping, Uppsala, Oslo, and Providence.

In particular, I am grateful to Gunnar Björnsson, Anders Tolland, Don Loeb, and Benoit Guilielmo for their insightful comments and helpful discussion.

A separate thanks goes out to David Christensen for generously inviting me to Brown University and working with me on the Level-Splitting view (although I know that he disagrees with me).

Thanks to the people at the Philosophy department at Brown for being friendly and welcoming. Thanks also to the Sweden-America Foundation for making it possible for me to go there.

A warm thanks to Maria Lasonen-Aarnio for providing extensive and very helpful comments on the entire manuscript for the dissertation at the end of the process.

I would also like to thank the people working at

the department for providing a very nice and

stimulating environment. It has been a pleasure to

work and have lunch (e.g., tuna sandwich) with

great people like Alva Stråge, Susanna Radovic, Per

Andersson, Erik Joelsson, Thomas Hartvigsson, Bengt Brülde, Moa Ekbom, Morten Sager, Leila El- Alti, Joakim Sandberg, Sofia Jeppson, Christian Munthe, Dorna Behdadi and many more. Thanks also to my roommate Alexander “golden goose”

Andersson for great company and for sharing my interest in country music.

Finally, I would like to thank friends and family for emotional support and encouragement during this period of years. To my amazing children, Amandine, Adriano, and Marguerite but especially to my beautiful wife Matilde for believing in me and showing patience with all of my eccentricities.

Marco Tiozzo, Gothenburg, November 2018.

I NTRODUCTION TO THE PAPERS ...1

1 Moral Disagreement...5

1.1 Moral Skepticism...5

1.2. Arguments for Moral Skepticism...8

1.3. Arguments from Disagreement...10

2 The Significance of Higher-Order Evidence....15

2.1 Higher-Order Evidence...16

2.2 A Puzzle...17

2.3 The Debate...19

3 The Argument from Peer Disagreement...25

3.1 Peer Disagreement...25

3.2 From Peer Disagreement to Moral Skepticism...30

3.3 Resisting the Argument...32

4 The Argument from Faultless Disagreement. .35 4.1 Faultless Disagreement...35

4.2 From Faultless Disagreement to Moral Skepticism...37

4.3 The Epistemic Account of Faultlessness...38

5 The Argument from Dogmatism...41

5.1 Moral Intransigence...42

5.2 From Moral Intransigence to Moral Skepticism...43

5.3 The Functional Role of belief...45

References...47

This Doctoral Thesis is a study of arguments based on considerations which have to do with moral disagreement. Arguments from disagreements can be used in order to argue for or against various metaethical positions. Most often, however, arguments from disagreement are used in an attempt to establish some sort of moral skepticism.

One might distinguish two general types of skeptical arguments from disagreement. The first type of argument purports to establish epistemological moral skepticism: skepticism about moral knowledge or moral justification, whereas the other type of argument purports to establish ontological moral skepticism: skepticism about moral reality.

In the dissertation, I consider an argument for

epistemological moral skepticism that is based on

considerations about peer disagreement – roughly,

a dispute in which both parties are in an equally

good epistemic position to reach the truth of the

relevant matter at issue. Peer disagreement and

higher-order evidence more generally is supposed

to provide a defeater for one’s belief about the

relevant matter at issue. The argument holds that

our controversial moral beliefs – which seem to be

most, if not all, of our moral beliefs – for this reason

do not amount to knowledge. In response to this

argument, I develop a new approach to the

normative significance of peer disagreement (and

higher-order evidence more generally). I argue that

higher-order defeat is best explained by the fact

relevant matter fails to be rational, rather than by the fact that one comes into possession of higher- order evidence which supports believing that one’s belief fails to be rational. An important consequence of this explanation is that higher-order defeat turns out to be a contingent matter. I argue that the condition higher-order defeat is contingent upon is very often not satisfied when it comes to moral peer disagreement specifically. As a result, it appears that moral knowledge is very seldom threatened by peer disagreement.

Together with John Eriksson I also consider two arguments from disagreement that purport to establish ontological moral skepticism. The first argument holds that the possibility of faultless disagreement in the moral domain is something that speaks in favor of some anti-realist alternative (e.g., expressivism or relativism) rather than moral realism. The second argument holds that the possibility of intransigence to moral disagreement and other sorts of putative counterevidence to one’s moral judgments is something that speaks against cognitivism.

The dissertation consists of this general introduction and the following five papers: “Higher- Order Defeat: An Explanatory Problem” (ms a);

“The Level-Splitting View and the Non-Akrasia Constraint” (Tiozzo forthcoming); “Moral Disagreement and the Limits of Higher-Order Defeat” (Tiozzo ms b); “Matters of Ambiguity:

Faultless Disagreement, Relativism and Realism”

(Eriksson and Tiozzo 2016); and “The Moral Dogmatist: An Argument Against Cognitivism”

(Eriksson and Tiozzo ms).

Two of the papers (Tiozzo ms a; forthcoming), do not directly address moral disagreement. They are, however, necessary in order to develop the technical machinery I use in the third paper (Tiozzo ms b), where I argue that certain limits to higher- order defeat mitigate the skeptical threat from moral disagreement. These three papers can also be read as independent arguments and views about the significance of higher-order evidence and peer disagreement. The two papers that are co-authored (Eriksson and Tiozzo 2016; ms) are less directly concerned with higher-order evidence and peer disagreement; instead they explore and develop arguments for moral skepticism that draw on other sorts of observations about moral disagreement.

While moral philosophers have been interested in considerations to do with disagreement for quite some time, epistemologists have only recently started to investigate the phenomenon more closely.

1

The most tangible sign of this is the recent debate over the epistemic significance of peer disagreement, and the impressive amount of works that have been published on the subject.

I think that there is much to be gained by examining issues regarding disagreement on an interdisciplinary philosophical level. Ethics and epistemology are closely related philosophical disciplines. Both disciplines are concerned with normative questions:

ethics with the assessment of actions and the general principles which govern them, and

1 Reflections on the significance of moral disagreement are prominent in classic works in metaethics such as Sidgwick (1874), Ayer (1936), Stevenson (1944), and Hare (1952).

2 See Matheson (2015) for a nice overview of some of the

most prominent positions and arguments in the debate

over the normative significance of peer disagreement.

epistemology with the assessment of doxastic attitudes and the general principles which govern them. I hope that this dissertation may contribute to some extent to bringing these disciplines closer together, at least on issues concerning the normative significance of disagreement.

The principal aim of this introduction is to provide a general background and clarify the philosophical contribution of the thesis. In order to do so, I will present the wider philosophical context in which the papers are situated and explicate in what way the arguments presented in the dissertation relate to and differ from other positions in the debate.

The introduction is divided into five sections. In section 1 I sort out various arguments for moral skepticism. The purpose is to clarify what is distinctive about arguments from disagreement. In section 2 I present the debate over the significance of higher-order evidence and explicate how the view I develop in the dissertation relates to other views in the debate. In sections 3, 4, and 5 I provide some background for the specific arguments from disagreement that are discussed in the dissertation.

Section 3 presents what I call the “Argument from Peer Disagreement”, Section 4 what I call the

“Argument from Faultless Disagreement”, and

Section 5 what I call the “Argument from

Dogmatism”.

Considerations having to do with moral disagreement have been used by skeptics since the dawn of philosophy. The extent and resilience of moral disagreement is supposed to make it unlikely that there are any objective moral facts, or at least that we could have anything like moral knowledge.

Philosophers disagree, however, about the extent to which arguments from disagreement provide a good case for moral skepticism. Some authors take moral disagreement very seriously. For instance, Don Loeb (1998) considers the argument from disagreement to be: “among the most important that has been directed against moral realism” (p.

281). While others, such as David Copp (1991), Russ Shafer-Landau (2003), and David Enoch (2009) have argued that arguments from disagreement do not obviously undermine, or even challenge, moral realism in any substantive way.

The aim of this first section is to provide a wider background for the skeptical arguments from disagreement that I discuss in the dissertation. I begin by defining moral skepticism and making some useful distinctions in section 1.1. Section 1.2 describes various types of arguments that have been used in order to make a case for moral skepticism. In section 1.3, I turn to arguments from disagreement in particular.

1.1 Moral Skepticism

The skeptic believes that for anything you think you

Descartes famously presented the hypothesis that we could all be the victims of a powerful and deceitful evil genius. This hypothesis explains how all of our beliefs could be false. A more up-to-date skeptical hypothesis states that we might all be brains in vats. Moral skepticism is different.

Arguments for moral skepticism do not normally proceed from some skeptical hypothesis.

Instead, they appeal to considerations which have to do with cases of real-world moral disagreement. In 1.2 I will briefly describe some different arguments for moral skepticism. One of the most common types of argument, and the one that will be the focus of this thesis, starts from moral disagreement.

I will use moral skepticism as a blanket term for a variety of skeptical claims about morality. As I mentioned at the outset, one variety of moral skepticism, epistemological moral skepticism, is skepticism about moral knowledge or moral justification, while another, ontological moral skepticism, is skepticism about moral reality.

Epistemological moral skepticism is the claim that we do not have any moral knowledge or that we do not have any justified moral beliefs. Although moral knowledge skepticism and moral justification skepticism are logically distinct it is normally assumed that moral justification skepticism entails

3 An exception is Walter Sinnott-Armstrong (2006) who

argues that we cannot have any moral knowledge since

there is “no way to rule out moral nihilism” (p. 79). Joshua

May (2013) provides some good arguments to show that it

is easier said than done to develop a moral analogue to the

skeptical hypothesis argument for perceptual skepticism.

moral knowledge skepticism.

One may also draw a distinction between Academic and Pyrrhonian versions of moral skepticism. Academic moral skepticism maintains that moral knowledge and moral justification are impossible, whereas Pyrrhonian moral skepticism maintains that one must suspend judgment over whether moral knowledge and moral justification are possible.

By contrast, arguments for ontological moral skepticism target moral realism. There are various conceptions of moral realism in the literature, but for our purposes we can understand the position broadly as the conjunction of: cognitivism; success- theory; absolutism; and objectivism.

Cognitivism is the claim that moral judgments express truth-apt beliefs. Success-theory is the claim that many moral propositions are true. Absolutism is the claim that the truth value of moral judgments is independent of individual or collective frameworks, i.e., the rejection of contextualism and truth-value relativism. Finally, objectivism is the claim that moral facts are mind-independent facts. Moral realism can, in turn, be either naturalistic or non- naturalistic. Moral naturalism is the view that

4 Notice, however, that in some epistemological views, knowledge does not require justification. See, e.g., Littlejohn (2012) and Sutton (2007).

5 This definition has the advantage of encompassing all, or at least most, arguments that are considered to speak against moral realism. However, people do sometimes have a weaker conception in mind when they speak about moral realism. “Minimal” moral realism is the view that moral propositions are truth-apt and that some of them are true.

One might also consider moral relativism, moral

constructivism, and the types of moral non-cognitivism that

endorse a deflationary account of truth to be forms of

moral realism.

moral properties are identical with, supervenes on, or are reducible to, natural properties, whereas moral non-naturalism is the view that moral properties are sui generis.

The two principal forms of ontological moral skepticism are moral error theory and moral non- cognitivism or “non-descriptivism”. Proponents of moral non-cognitivism do not take moral judgments to be truth-apt. As a consequence, they do not normally think that there are any objective moral facts. In contrast, proponents of moral error theory hold that moral judgments are truth-apt. Moral error-theory is still a form of ontological moral skepticism as it maintains that all moral judgments are false. Proponents of moral error-theory claim that when we make moral judgments we misrepresent the world.

Notice that there is a close connection between epistemological and ontological moral skepticism.

First, arguments for ontological moral skepticism are often arguments for epistemological moral skepticism. Without any moral beliefs or moral truths we cannot have moral knowledge given the uncontroversial assumption that knowledge requires true belief. Second, although skepticism about moral knowledge or moral justification is logically consistent with the existence of moral facts, it is seems that epistemological moral skepticism would significantly weaken the case for moral realism.

1.2. Arguments for Moral Skepticism

A variety of arguments have been presented

in order to attempt to establish either

epistemological or ontological moral

skepticism. Diego E. Machuca (2017) distinguishes four main types of arguments:

the “argument from the best explanation”;

the “argument from queerness”; the

“argument from evolution”; and the

“argument from disagreement”.

Here is a brief recap.

The argument from the best explanation is discussed particularly by Gilbert Harman (1977) and holds that our having moral beliefs is better explained by psychological and socio-cultural facts about us, than by the existence of objective moral facts. Since this makes objective moral facts explanatorily redundant we should assume that there are no such facts, i.e., ontological moral skepticism. In response, Nicholas Sturgeon (1988) has argued that the argument only will convince someone who already accepts moral skepticism.

Machuca (2017, p. 14) points out that one might also argue from the argument from the best explanation to the weaker thesis that if objective moral facts are explanatorily redundant, then one does not have any reason to believe in their existence. In that case, the argument implies epistemological moral skepticism rather than ontological moral skepticism.

The argument from queerness identifies ways in which moral properties and facts are queer. It was

6 There are also other arguments for moral skepticism that

do not fit within these four general types of argument. For

example, one type of argument holds that the motivating

component of moral judgments is something that speaks in

favor of non-cognitivism rather than cognitivism (see

Björnsson et al. 2015). I think, nevertheless, that these four

types of arguments are the most well-known and frequently

discussed ones in the literature.

first developed by J. L. Mackie (1977). One might formulate the argument in different ways depending on which queer element one is targeting: the motivating or normative nature of allegedly objective moral facts, the supervenience of these facts on natural properties, or the knowledge of objective moral facts that we are supposed to have. Depending on how one frames it, the argument from queerness can be used either to establish ontological or epistemological moral skepticism. In response to the argument from queerness one might deny that moral properties are queer in the relevant sense or argue that the argument overgeneralizes since many properties share the alleged queer features.

However, as Jonas Olson (2014) points out, the strength of such objections will depend on what version of the argument from queerness one has in mind.

The argument from evolution maintains that the most plausible account of the origin of morality is one that appeals to evolution. A crucial part of this account is that biological evolution generates moral belief-forming processes that are adaptive, rather than moral belief-forming processes that are reliable. The argument from evolution can be used to argue for both epistemological and ontological moral skepticism. Debunking arguments of this sort are most often used to argue that our moral beliefs fail to be justified since their source is not trustworthy.

But sometimes the argument from evolution is also used in order to argue directly against moral realism.

A popular response to the

7 See Olson (2014, Ch. 5 and 6) for an in-depth evaluation of the argument from queerness.

8 See, e.g., Joyce (2001).

9 See, e.g., Street (2006).

argument from evolution is to propose a so-called

“third-factor” that is supposed to explain how our moral judgments could be correlated with the moral truths even if the moral truths do not explain our moral beliefs.

The debate has continued and at this stage there is a burgeoning number of works on evolutionary arguments.

Finally, there is the argument from disagreement, which holds that considerations which have to do with disagreement can be used to establish either epistemological or ontological moral skepticism.

Next, we will take a closer look at different versions of this type of argument.

1.3. Arguments from Disagreement

Epistemological arguments from disagreement purport to establish that we do not have any moral knowledge or moral justification. This type of skeptical argument normally grants that there are objective moral facts. The aim of epistemological arguments is instead to show that a certain type of moral disagreement will defeat moral knowledge or moral justification. In the following I will take it that two persons genuinely disagree over a proposition iff they hold incompatible doxastic attitudes toward that proposition.

An argument for epistemological moral skepticism that has recently received much attention is based on observations of peer disagreement. A peer disagreement is, broadly, a dispute in which both parties are in an equally good epistemic position in relation to finding out the

10 Enoch (2010) was among the first to provide an argument to this affect.

11 For a nice overview see Wielenberg (2016).

truth of the relevant matter at issue. According to a widely embraced view about the significance of peer disagreement this should make the disputants suspend judgment about the disputed matter. Since peer disagreement is more or less routine in the moral domain it has been argued that peer disagreement therefore leads to moral skepticism.

Arguments from peer disagreement can either be framed in terms of moral knowledge skepticism or moral justification skepticism. When Sarah McGrath (2008) formulated the argument from peer disagreement it was only supposed to be an argument for skepticism about moral knowledge.

McGrath’s version of the argument leaves it open as to what extent moral beliefs can be justified in the face of peer disagreement. Other formulations of the argument are instead supposed to be arguments for skepticism about justified moral belief. For example, Katia Vavova (2014) presents the argument from peer disagreement as an argument for agnosticism about morality. However, since it is normally assumed that knowledge requires justified belief, the argument thereby also lends support to moral knowledge skepticism.

A common form of ontological argument from

disagreement starts out with the observation that

there are widespread and persistent moral

disagreements that appear resistant to rational

resolution. It is then argued that this type of moral

disagreement, for some reason or the other, is

incompatible with moral realism. According to a

well-known version of this argument, moral realism

fails to provide the best explanation for widespread

and persistent moral disagreement. Mackie (1977)

famously argued that moral disagreement is better

explained by the hypothesis that moral opinions reflect socio-cultural factors, than by the hypothesis that there are objective moral facts that some cultures or persons have access to while others do not. Mackie’s argument has been much criticized but I will set these criticisms aside for the purpose of this dissertation.

Another way to state the argument is in terms of epistemic access.

The main assumption of this version of the argument is that some moral disagreements are not even in principle resolvable;

they are supposed to be radical in the sense that they cannot be attributed to cognitive shortcomings. The fact that such moral disagreements are not even in principle resolvable is in turn supposed to indicate that moral facts are unknowable. Since the plausibility of moral realism depends on the assumption that we have access to morality, it seems to follow that moral realism is false. As Loeb (1998) points out: “The plausibility of moral realism depends on the assumption that we have direct (non-inferential) access to morality” (p.

282). In response, moral realists have provided so- called defusing explanations to argue that most, if not all, moral disagreements can be explained away in terms of cognitive shortcomings and differences in background beliefs.

12 See, e.g., Adams (2008), Brink (1989), McGrath (2008), and Tersman (2006) for various objections to Mackie’s argument.

13 See, e.g., Bennigson (1996), Loeb (1998), Tersman (2006), and Tolhurst (1987) for interpretations of the argument from disagreement in terms of (lack of) epistemic access.

14 I borrow the expression “defusing explanations” from

Doris and Plakias (2008). See, e.g., Brink (1989) for a wide

Another type of ontological argument from disagreement draws on intuitions about disagreement to reach a semantic conclusion.

Non-cognitivists hold that moral disagreement is to be explained in terms of conflicting attitudes, whereas moral realists hold that moral disagreement is to be explained in terms of conflicting beliefs. However, it seems that some moral disagreements do not have to involve conflicts of beliefs. In a famous passage of The Language of Morals, R. M. Hare asks us to imagine a missionary arriving at a faraway tropical island inhabited by cannibals. The missionary and the cannibals soon engage in debate over the moral status of hunting scalps. The cannibals consider scalp hunting to be morally good and the missionary finds it reprehensible. The twist of the story, however, is that the cannibals and the missionary are using the moral term “good” in different ways. When the cannibals use the word

“good” they are applying it to persons who are

“bold and burly and collect more scalps than the average”, whereas the missionary applies “good” to persons who are “meek and gentle and do not collect large quantities of scalps” (1952, p. 148 ff.).

The story is supposed to show that it is possible to have a moral disagreement without using moral terms to denote the same things. This, in turn, is supposed to indicate that moral disagreement is not to be explained in terms of conflicting beliefs but

array of defusing explanations.

15 Folke Tersman (2006) calls this the “Argument from

Ambiguity”. See, e.g., Blackburn (1984), Hare (1952),

Horgan and Timmons (1990), Loeb (1998), Stevenson

(1944), and Smith (1994) for various versions of this

argument.

rather in terms of conflicting non-doxastic attitudes (e.g., desires). Notice that the conclusion of the argument is not that moral properties do not exist, but rather that there is no single moral property to denote moral terms like “rightness”, which in turn does not square very well with the explanation of moral disagreement favored by moral realists. In response, moral realists (e.g., Boyd 1988) have appealed to a causal theory of reference that is supposed to be able to account for moral disagreement in terms of conflicting beliefs.

In Tiozzo (ms b) I target an epistemological argument from disagreement. The argument proceeds from peer disagreement and purports to establish moral knowledge skepticism. By contrast, both arguments that Eriksson and I discuss are, in different ways, ontological arguments from disagreement. The argument that we examine in Eriksson and Tiozzo (2016) draws on the possibility of faultless moral disagreement  a phenomenon that seems difficult to accommodate within the moral realist framework. In Eriksson and Tiozzo (ms) we introduce another type of argument for ontological moral skepticism. This argument (which is related to an argument from disagreement formulated by Mark Eli Kalderon (2006)) is based on the possibility of the existence of a moral dogmatist, someone who makes a moral judgment and is not disposed to revise that judgment in the face of putative counterevidence, such as per disagreement.

16 Terence Horgan and Mark Timmons have in a series of papers about “moral twin earth” developed a version of the argument that also targets the causal theory of reference.

See Plunkett and Sundell (2013) and Dowell (2015) for

objections against this type of argument.

Before I go on to more closely present the

arguments from disagreement that are discussed in

this dissertation, I want to introduce the debate

over the significance of higher-order evidence. This

debate is crucial to understanding arguments based

on peer disagreement.

Evidence

Generally speaking, any factor that makes it more probable that certain states of affairs obtain (or do not obtain) is evidence (that those states of affairs obtain). Such factors might include beliefs, memories, facts, propositions, experiences, events, and worldly objects. For example: the presence of dark clouds is evidence of rain; observations about the correlation between smoking and cancer are evidence that smoking causes cancer, and so on.

According to a widely embraced view in epistemology, known as evidentialism, one is rationally required to believe in accordance with one’s evidence. For instance, if one’s evidence supports believing that “it is going to rain shortly”, then one is rationally required to believe that “it is going to rain shortly”. More generally, if one’s evidence supports believing p, then one is rationally required to believe p according to evidentialism.

However, the relationship between evidence and

rational belief becomes more complicated when the

evidence concerns the rationality of one’s belief

about p at a meta-level. This is especially true in

cases where the evidence about the epistemic

status of one’s belief is misleading. For example, it

seems possible to have evidence to believe that

one’s belief about p fails to be rational, although

one’s belief about p is in fact rational. In cases like

this it is not obvious what one should believe given

one’s total evidence.

higher-order evidence. First, in section 2.1, I identify and discuss some features that I take to be distinctive of higher-order evidence. Then, in section 2.2, I present a puzzle that arises in relation to the evaluation of the evidential impact of higher- order evidence. Finally, in section 2.3, I present different positions in the debate and explicate how the view I develop about the normative significance of higher-order evidence relates to these other views.

2.1 Higher-Order Evidence

In epistemology nowadays, it is common to separate “first-order” evidence from “second- order”, or “higher-order” evidence. First-order evidence is evidence about what to believe about some object-level proposition p (e.g., it is going to rain shortly), whereas higher-order evidence is evidence about what to believe about one’s belief about p. More precisely, higher-order evidence is evidence about the epistemic status of one’s belief about p (e.g., it is rational for me to believe that it is going to rain shortly, or my evidence supports believing that it is going to rain shortly).

First-order evidence bears directly on what to

believe about the relevant matter. For example, the

presence of dark clouds constitutes first-order

evidence that supports believing that it is going to

rain shortly, and observations about the correlation

between smoking and cancer constitute first-order

evidence that supports believing that smoking

causes cancer, and so on. By contrast, higher-order

evidence bears directly on what to believe about

the epistemic status of one’s belief about p but also, according to many authors, indirectly on what to believe about p.

It is therefore often assumed that higher-order evidence has the ability to defeat one’s belief about p, i.e., to make it less rational or justified.

Cases that are intended to elicit intuitions about higher-order defeat involve sleep deprivation, mind- distorting drugs, biases of various sorts, and peer disagreement. For example, learning that someone has slipped a very powerful hallucinatory drug in one’s morning coffee should make one revise one’s belief that there is a pink flamingo at the breakfast table. Sufficiently strong higher-order evidence therefore appears to have the ability to defeat one’s belief about the relevant matter at issue. However, as some authors have noted, higher-order defeat appears to be very different from familiar types of defeat.

First, unlike rebutting defeaters, higher-order defeaters do not typically provide one with evidence to believe that the relevant proposition is false.

Rather, what higher-order evidence implies is that one does not know whether one has reasoned

17 The distinction between direct and indirect evidence might be difficult to uphold upon closer inspection, at least if indirect evidence is supposed to be evidence that depends on some further assumption. It seems to me that most evidence requires some background assumption in order to support believing a certain proposition. Thanks to Don Loeb for pointing this out to me.

18 Advocates of higher-order defeat include: Christensen (2010), Feldman (2009), Horowitz (2014), Kelly (2010), Matheson (2009), Vahid (2015), Schechter (2013), and Sliwa and Horowitz (2015).

19 See, e.g., Christensen (2010), DiPaolo (2016), and

Lasonen-Aarnio (2014).

correctly from the evidence. In this sense, higher- order defeaters are similar to undercutting defeaters. However, unlike undercutting defeaters, higher-order defeaters have a retrospective aspect.

Higher-order defeaters give one reason to believe that one’s belief was never rational to start with. This is different to how undercutting defeaters work, since they do not entail that one’s belief failed to be rational given the original evidence.

Finally, and most importantly, given that one’s belief  at least initially  reflects a rational response to the evidence, we have to presume that the higher-order evidence is misleading (Tal forthcoming). In one sense, this does not change much since misleading evidence is evidence nonetheless.

However, as a result of the misleading nature of higher-order evidence, it seems that higher-order evidence might leave evidential support intact (Christensen 2010, p.

195).

The peculiar nature of higher-order evidence creates an enigmatic puzzle about how to evaluate one’s total evidence. Next, we will take a closer look at this puzzle.

2.2 A Puzzle

Richard Feldman (2005) was the first to draw attention to a certain epistemological puzzle that arises in relation to misleading higher-order

20 This expression comes from Lasonen-Aarnio (2014).

21 Authors like Smithies (2012) and Titelbaum (2015)

appear to think that it is impossible to have misleading

higher-order evidence of this sort.

evidence.

To introduce the puzzle, here is an abbreviated version of a case from Sophie Horowitz (2014):

Sleepy Detective. Sam is a detective working to identify a jewel thief. Staying up the whole night and considering all the available evidence (codes, photographs, letters etc.), he correctly reasons to the conclusion that Lucy is the thief. He then informs his partner, Alex, about his findings. Alex responds by reminding him that in the last ten cases where he has spent the whole night assessing the relevant evidence, his reasoning has been unreliable. So, Alex concludes that there is no reason to think that Sam’s evidence supports Lucy being the thief in this case. Sam rationally trusts Alex and believes that she is right.

Two questions arise: what should Sam believe about who the jewel thief is? And what should Sam believe about what his evidence supports? Horowitz describes Sam’s epistemic position in the following way: first, Sam’s first-order evidence, E, which includes codes, letters, and photographs supports believing p, that Lucy is the thief. Second, Sam’s higher-order evidence, which includes Alex’s testimony about his track record data, supports believing that E does not support p. The crucial question then becomes: what does Sam’s total evidence support?

Horowitz points out that the two answers that have been most commonly defended in the literature are the “Conciliatory” view and the

22 Puzzles very similar to Feldman’s are also presented in Horowitz (2014) and Sliwa and Horowitz (2015).

23 This is an adaptation of a case that appears in Horowitz

(2014).

“Steadfast” view.

According to the Conciliatory view, Sam should become much less confident in p, and much less confident that E supports p.

According to the Steadfast view, Sam should maintain his high confidence in p, and also his high confidence that E supports p. Notice that both conciliatory and steadfast views assume that there is a connection between what one is rational to believe and what one is rational to believe one is rational to believe. In this sense, they are both

“level-connecting” views.

A third alternative is to split the levels and accept divergent rational recommendations at each level.

That is, to accept that what one is rational to believe does not have to be the same as what one is rational to believe one is rational to believe.

According to the “Level-Splitting” view, Sam should maintain high confidence in p, but become much less confident that E supports p.

Many people find this alternative intuitively implausible (more about why in the next section).

What makes this an epistemological puzzle is that all alternatives appear to be bad in some way. The Conciliatory view respects the higher-order evidence but fails to respect the first-order evidence. The Steadfast view respects the first- order evidence but fails to respect the higher-order evidence. The Level-Splitting view respects both

24 These terms are most often used to denote the main positions in the peer disagreement debate. Horowitz also uses them in a more general sense to denote different views about the normative significance of higher-order evidence. I will follow suit.

25 Philosophers who have advocated the Level-Splitting

view include Coates (2012), Lasonen-Aarnio (2014), and

Weatherson (ms).

the higher-order and the first-order evidence but appears to be intuitively implausible.

This puzzle has launched a wide-ranging debate over the normative significance of higher-order evidence. Next, we will take a closer look at this debate.

2.3 The Debate

The debate over the normative significance of higher-order evidence is complex. On the one hand, there is a conflict between conciliatory views and steadfast views about whether evidential support primarily trickles down or bubbles up. On the other hand, there is a conflict between level-connecting views and level-splitting views over whether there is a connection between what one is rational to believe at different epistemic levels.

First, let us take a look at the conflict between conciliatory and steadfast views. Proponents of the Conciliatory view hold that it is the higher-order evidence about p that primarily determines what one ought to believe about p, whereas proponents of the Steadfast view hold that it is the first-order evidence about p that primarily determines what one ought to believe about p.

Thomas Kelly (2010) suggests that both conciliatory and steadfast views are on to something. The Conciliatory view proceeds from a genuine insight he calls downward epistemic push:

“what it is reasonable for one to believe about the world is not wholly independent of what it is reasonable for one to believe about what it is reasonable for one to believe about the world” (p.

158). However, Kelly also points out that there is

another insight to balance the one about downward

epistemic push, which he calls upward epistemic push: “when E is genuinely good evidence for H, this very fact will contribute to the justification for believing the epistemic proposition that E is good evidence for H that is available for those with the relevant competence” (p. 159). This type of insight appears to speak in favor of the Steadfast view.

Kelly (2010) goes on to defend an ecumenical view he calls the “Total Evidence” view. According to this view, the evaluation of the total evidence will depend on the relative strength of the higher-order and first-order evidence. He predicts that in some cases, the higher-order evidence will swamp the first-order evidence and one will be rationally required to give up one’s belief. In other cases, it will be the other way around; the first-order evidence will swamp the higher-order evidence and one will be rationally permitted to retain one’s belief.

Although, Kelly’s proposal appears to have the advantages of both conciliatory and steadfast views, it also appears to have the disadvantages of both views. By aggregating the evidence to find a suitable compromise, it seems that one will neither give the higher-order evidence nor the first-order evidence its full due.

The other debate over the significance of higher-

order evidence regards the connection between

epistemic levels. Although proponents of

conciliatory and steadfast views disagree on

whether evidential support trickles down or

bubbles up, they agree that there is a close

connection between what one is rational to believe

about p and what one is rational to believe one is

rational to believe about p. In this sense they are

both level-connecting views. By contrast, proponents of level-splitting views do not think that evidential support either has to trickle down or bubble up. Instead they hold that the total evidence sometimes supports divergent beliefs at different epistemic levels, i.e., that one’s total evidence can be self-misleading.

The main motivation for level-connection views comes from considerations having to do with coherence. There seems to be something absurd or Moore-paradoxical about believing, asserting, and acting upon akratic combinations of doxastic attitudes.

Some authors (e.g., Feldman 2005, Melis 2016, Smithies 2012) therefore argue that it is impossible for one’s total evidence to support akratic combinations of doxastic attitudes. Others (e.g., Christensen 2010, Horowitz 2014) seem to accept the possibility that one’s total evidence can be “self-misleading” in this sense and emphasize instead the fact that it is always irrational to occupy an akratic combination of attitudes.

Consider Sleepy Detective again. Horowitz (2014) thinks that it is intuitively odd for Sam to simultaneously believe that Lucy is the thief and that the evidence does not support that Lucy is the thief. Moreover, she argues that the oddness extends well beyond the oddness of the

26 Richard Feldman finds it “extremely odd” to make claims like p, but my overall evidence does not support p (2005, p.

108). Much in the same spirit Michael Bergmann states

that “it is an epistemically bad state of affairs if I believe

both that P and that my belief that P is not formed in a

trustworthy way.” (2005, p. 424) Others, such as Adler

(2002), go further and argue that epistemic akrasia is

impossible. For the sake of argument, I will assume that

epistemic akrasia is possible.

combination of attitudes themselves. For instance, what would happen if Sam were to act on his akratic beliefs. Someone who is convinced that Lucy is the thief should be disposed to recommend that the cops raid her apartment and send her to jail. But if Sam were asked to justify or explain his behavior he would not have much to say since he does not think that his evidence supports any of it.

Horowitz also points out that akratic combinations of attitudes seem to vindicate irrational behavior.

She observes that: “it seems patently irrational to treat a bet about P and a bet about whether one’s evidence supports P as completely separate”

(Horowitz 2014, p. 728).

In contrast, advocates of the Level-Splitting view accept that misleading higher-order evidence can sometimes make one rationally required to take an akratic combination of beliefs. Allen Coates (2012) discusses an example using Sherlock Holmes that is structurally similar to Sleepy Detective. Holmes brings Watson to a crime scene. Watson gets the opportunity to evaluate the evidence by himself and reasons correctly to the conclusion that the evidence supports believing that the butler is guilty. But when Watson shares his conclusion with Holmes, Holmes tells him that his conclusion is irrational. Since Holmes is a master detective it seems rational for Watson to believe that the evidence does not support believing that the butler is guilty. Coates concludes that: “both Watson’s belief and his epistemic judgment of it are rational, even though together they constitute epistemic akrasia” (p. 122).

As is apparent, proponents of the Level-Splitting

view and the Level-Connection view reason about

these type of cases in very different ways. Several authors have recently pointed out that evidential reasons and coherence requirements should be treated separately.

In Tiozzo (forthcoming) I argue, in a similar vein, that participants of the debate might be conflating two distinct conceptions of epistemic rationality. According to a substantive conception of epistemic rationality, one is rationally required to believe in accordance with one’s total evidence. By contrast, according to a structural conception of epistemic rationality, one is rationally required to make one’s attitudes coherent.

As a key example to illustrate my point, I use Horowitz’s argument against the Level-Splitting view. I argue that the appeal to epistemic akrasia is a non-starter in an argument directed against a view that presupposes a substantive conception of rationality (such as the Level-Splitting view). So, more generally, whether or not one endorses higher-order defeat might depend on what conception of rationality one has in mind.

Moreover, and more importantly, as I argue in Tiozzo (ms a), higher-order defeat only appears plausible given a structural conception of epistemic rationality. I also point out that a very interesting consequence of this is that higher-order defeat becomes a contingent matter. As I will explain in the following section, this opens up the possibility of mitigating the skeptical challenge from arguments based on peer disagreement.

I will spend the rest of this general introduction presenting the three arguments from disagreement

27 Most notably Lasonen-Aarnio (forthcoming) and Worsnip

(2018).

that figure in the dissertation. First up is the

Argument from Peer Disagreement.

Disagreement

The Argument from Peer Disagreement has emerged from the wake of the debate over the epistemic significance of peer disagreement. The upshot of the argument is that peer disagreement ensures that many of most people’s moral beliefs do not amount to knowledge. McGrath’s “Moral Disagreement and Moral Expertise” (2008) is the locus classicus of this type of argument and has triggered much discussion. Similar arguments has also been formulated in order to argue for skepticism in other epistemic domains in which peer disagreement seems to be a common phenomenon, such as religion or philosophy in general.

In Tiozzo (ms b) I argue that this argument fails to establish moral skepticism. The main reason for this, I argue, is that peer disagreement (and higher- order evidence more generally) only contingently gives rise to defeat, and importantly, that the condition it is contingent upon is very often not satisfied when it comes to moral peer disagreement specifically.

The purpose of this part is to provide some

background to the Argument from Peer

Disagreement. In section 3.1 I introduce the debate

over the normative significance of peer

disagreement. I will also say something about what

motivates conciliatory and steadfast views

respectively. Section 3.2 explains how one can go

of resisting the Argument from Peer Disagreement and explicate how my argument differs from the other arguments.

3.1 Peer Disagreement

People disagree about virtually everything.

Sometimes it is fairly obvious that one’s opponent is mistaken. For example, in situations in which I am significantly more competent in the relevant area of disagreement, or when I have access to crucial evidence that the other party lacks, it appears that it is my opponent  rather than me – who is mistaken. At other times, however, I might have no more reason to suspect that the other party is in error than I am. In these cases, we appear to be in a peer disagreement.

“Epistemic peers” is a technical term that has been characterized in various different ways in the literature.

To be epistemic peers in relation to a certain matter is, roughly, to be epistemically on a par with respect to finding out the truth of the disputed matter at issue. According to an influential suggestion by Kelly (2005) this means that the disputants have equal cognitive capacities and have considered the relevant body of evidence equally carefully. Relevant cognitive capacities include intellectual virtues such as intelligence, thoughtfulness, competence, and freedom from bias etc. Notice that it is important that the disputants share the same body of evidence, or at least that they have evidence that is on a par: otherwise the disagreement could easily be explained away by the

28 See Gelfert (2011) for an overview.

fact that either one disputant lacks access to an important piece of the evidence.

Here is a much-quoted example of peer disagreement from David Christensen:

Bill Calculation. Suppose that five of us go out to dinner. It's time to pay the check, so the question we're interested in is how much we each owe. We can all see the bill total clearly, we all agree to give a 20 percent tip, and we further agree to split the whole cost evenly, not worrying over who asked for imported water, or skipped desert, or drank more of the wine. I do the math in my head and become highly confident that our shares are $43 each.

Meanwhile, my friend does the math in her head and becomes highly confident that our shares are $45 each. How should I react, upon learning of her belief? (2007, p. 193)

Advocates of the Conciliatory view claim that one should be significantly less confident about one’s belief in the face of recognized peer disagreement, such as the case above. The most well-known version of conciliationism is the “Equal Weight”

view. According to this view one should give equal weight to the opinion of one’s peer and to one’s own opinion – which is usually interpreted as a reason to “split the difference”. The required doxastic revision could be cashed out either with the help of a more coarse-grained account (belief, disbelief, or suspension of judgment) or with the help of a more fine-grained account in terms of degrees of belief. As I see it, these are two alternative ways of describing the same thing and not two distinct epistemic entities. However, in order to facilitate the presentation, I will use the coarse-grained account.

The Conciliatory view is attractive since it

delivers the intuitive response to various

hypothetical cases of peer disagreement (e.g., Bill Calculation). It is also supported by theoretical considerations such as “Independence”. Here is one of Christensen’s formulations of the Independence principle:

Independence: In evaluating the epistemic credentials of another person’s belief about P, to determine how (if at all) to modify one’s own belief about P, one should do so in a way that is independent of the reasoning behind one’s own initial belief about P. (2009, p. 758)

According to Christensen, Independence is:

“intended to prevent blatantly question-begging dismissals of the evidence provided by the disagreement of others.” (2011, p. 2). There seems to be something wrong with disregarding a disputant’s opinion merely because oneself holds a different view oneself. So, given Independence it seems that one has nothing else to go on than the fact that a peer believes differently, which should motivate suspension of judgment, or should at least mean that one has significantly less confidence in one’s view regarding the disputed matter.

Christensen believes that this principle is what separates the camps in the peer disagreement debate (Christensen 2009, 2011).

29 For objections to Independence see, e.g., Kelly (2013)

and Lord (2014). A recent defense of Independence can be

found in Moon (2018).

Some authors (e.g., Feldman 2007) think that the Conciliatory view entails “Uniqueness”.

Here is a formulation from Roger White:

Uniqueness. Given one’s total evidence, there is a unique rational doxastic attitude that one can take to any proposition. (2005, p. 445)

Given that something like Uniqueness is correct, it seems that there can be no reasonable disagreements in the sense that both disputants hold rational attitudes regarding the disputed proposition. At most one of the disputant’s attitudes to the proposition can be rational.

By contrast, advocates of steadfast views tend to reject both Independence and Uniqueness.

According to the Steadfast view, one can be rationally permitted to retain one’s view despite evidence of peer disagreement. Steadfast views might be construed and motivated in different ways. Kelly (2010) suggests that we should make a general distinction between symmetrical and asymmetrical versions of the Steadfast view.

According to the symmetrical Steadfast view (e.g., Rosen 2001, Foley 2001, Wedgwood 2010), both disputants are rationally permitted to retain their views regarding the disputed matter, whereas

30 Feldman (2006), Kelly (2010), Ballantyne and Coffman (2012), Cohen (2013), and Schoenfield (2013) have argued that there is a tight relationship between versions of Uniqueness and conciliatory views about disagreement.

While others, such as Christensen (2009), have disputed the relationship between Uniqueness and conciliatory views.

31 Although this formulation might seem straightforward,

Matthew Kopec and Michael Titelbaum (2016) argue that

the label “Uniqueness” disguises a quagmire of different

versions of the thesis.

according to the asymmetrical Steadfast view (e.g., Kelly 2005, Lackey 2010, Titelbaum 2015), one (but not both) disputant is rationally permitted to retain her view regarding the disputed matter.

One way to motivate symmetrical steadfast views is to appeal to some type of epistemic permissiveness. Either by arguing that it is possible for epistemic peers to have reasonable disagreements, given that the shared body of evidence is underdetermined (Moffett 2007) or by arguing that it is possible to have reasonable disagreements given that the disputants are justified in following different epistemic norms (Goldman 2010). Both lines of argument appear to stand in conflict with Uniqueness.

It might be argued against permissive views that they only explain how synchronic peer disagreement is possible and not how diachronic peer disagreement is possible. The discovery of peer disagreement is supposed to provide new defeating evidence to the disputants. So, even if both disputants are permitted to take divergent doxastic attitudes to the relevant matter before the disagreement is revealed, it does not follow that they are also permitted to do so once the disagreement is out in the open.

Another way to try to motivate a symmetrical

Steadfast view is to argue that each disputant from

her own perspective is in the right to believe that

she has made a better evaluation of the shared

body of evidence. One might also use this type of

consideration to defend an asymmetrical steadfast

view by arguing that either disputant might be

justified in believing that her view is superior (e.g.,

Lackey 2010). However, without further

qualification this appears to be epistemic chauvinism. In order to argue for this type of steadfast view some authors have appealed to epistemic conservatism (Moffett 2007, Wedgwood 2010), others to self-trust (Foley 2001), and still others to the significance of private evidence (van Inwagen 1996, Lackey 2010).

The main problem with the first-person perspective is that there should be symmetry also from the first-person perspective. Christensen (2007) and Rattan (2014) argue that once the disputants reflect on the matter they should come to realize that the same factor might apply to the other party as well. Another problem, that Vahid (2014) brings to the fore, is that the appeal to the first-person perspective risks being question- begging. Writers who use this type of consideration presume from the start that evidence of peer disagreement should not affect one’s belief, which seems to be what needs to be established rather than merely presumed.

A third way to motivate steadfast views is to reject Independence and argue that both peers are rationally permitted to demote the other peer (Audi 2008, Bergmann 2009, Enoch 2010). For instance, according to Michael Bergmann (2009) it is possible for epistemic peers to rationally disagree since either peer might have an error-theory about why the other party is mistaken that mitigates the other peer’s judgment. On the other hand, this makes it harder to see how the dispute could qualify as a peer disagreement in the first place.

A fourth way to defend steadfastness is by

arguing that peer disagreement is of no substantive

epistemic significance. For instance, Kelly (2005)