Hannah Pillin

This dissertation develops a novel account of fragmented epistemic justification. It departs from and expands on the literature on fragmented accounts of belief, and develops a theory where not belief, but the justification of belief, is fragmented. This account is novel, and an addition to the already existing literature on fragmented accounts of belief. It will also shed light on and help to develop answers to some underdiscussed and unsolved questions in the existing literature on fragmentation, such as questions about the rationality of fragmentation. In a nutshell, what it means for justification to be fragmented will be that some piece of evidence might justify a certain proposition relative to some mental fragment, whereas that same piece of evidence might not justify that same proposition relative to a different mental fragment. The epistemic and non-epistemic circumstances, such as differences in accessible information, or different perspectives, focus, or practical goals etc., of the agent can vary across fragments, which is why these circumstances will be the parameters distinguishing fragments from each other. What exactly this means will become clear throughout this dissertation. For the first qualitative part of the dissertation, justification will be modelled to be a non-monotonic justification relation that holds between pieces of evidence and the propositions they justify, which will be indexed to fragments. Fragments, here, will be modelled by using total plausibility pre-orders over possible worlds. Later on, within a quantitative account of fragmented justification, there will be a variety of ways to spell out the concept of justification. Here, fragments will be modelled by using (subjective) probability functions or credence functions. After an introductory chapter, there will be a chapter focusing on an overview over relevant literature that is being discussed or worked with in this dissertation. Then, chapter 3 presents a skeleton of a qualitative (non-probabilistic) account of fragmented justification and shows how this account is able to provide a novel and attractive solution to the preface paradox (Makinson, 1965), one of the central paradoxes of epistemology. Then, chapter 4 develops a more full fledged account of qualitative fragmented justification, using this time not the preface paradox but a scenario from climate science as a motivating example for the account. Here, a scientific expert about climate science seems to be justified in believing an overall inconsistent set of propositions. The chapter explains how the toolkit of orthodox epistemology is limited in dealing with situations like this, and argues that fragmentation is a more appropriate strategy than existing approaches to deal with (seemingly rational) inconsistent doxastic states. The chapter provides a formalisation of this example, as well as a number of justifying reasons for why fragmentation can be rational. It, also, develops some precautionary norms of fragmentation. It concludes by explaining how fragmented justification relates to belief, and argues that fragmented justification provides a novel rationality argument for accounts of fragmented belief, and, furthermore, how one should go about de-fragmenting in light of practical rationality. Chapters 5 and 6 turn to a probabilistic or quantitative account of fragmented justification -- fragmented Bayesianism. Chapter 5 presents a very general version of fragmented Bayesianism, much more general than the existing accounts of fragmented Bayesianism that already exist. The basic idea is that the doxastic state of an agent will be represented not by one credence function, but by a set of credence functions, corresponding to different fragments. The focal point of this chapter is a list of parameters, such as background knowledge, algebra, confirmation measure, or ur-prior, that are part of the orthodox Bayesianism machinery, which, each, can be chosen to be either fragmented or global within a fragmented account of Bayesianism. Depending on which and how many of those parameters are chosen to be kept global or to render fragmented, the resulting account of fragmented Bayesianism will be more or less severely fragmented. The chapter provides motivations and arguments for why Bayesianism in general, and each of the parameters in particular, deserve fragmented versions. It also provides a brief comparison between the qualitative account of fragmented justification developed in the previous chapters with fragmented Bayesianism. Lastly, chapter 6 compares fragmented Bayesianism with another framework, imprecise Bayesianism. Imprecise Bayesianism aims to model agents who do not have enough information to form precise credences. Fragmentation, on the other hand, aims to represent agents with fragmented doxastic states, where fragments are individuated via differences in the agent’s epistemic and non-epistemic circumstances. The formal models underlying these two accounts are strikingly similar. In both frameworks, the agent's doxastic state is represented not via one but via many different credence functions. The literature on fragmentation and imprecision do not comment on the striking similarity between these two different accounts. This is why chapter 6 is devoted to presenting and comparing their similarities and differences. The dissertation closes with a conclusion and some remarks on potential avenues for further work.

This doctoral dissertation develops a novel account of fragmented epistemic justification. It departs from and expands on the literature on fragmented accounts of belief, and develops a theory where not belief, but the justification of belief, is fragmented. This account is novel, and an addition to the already existing literature on fragmented accounts of belief. It will also shed light on and help to develop answers to some underdiscussed and unsolved questions in the existing literature on fragmentation, such as questions about the rationality of fragmentation. In a nutshell, what it means for justification to be fragmented will be that some piece of evidence might justify a certain proposition relative to some mental fragment, whereas that same piece of evidence might not justify that same proposition relative to a different mental fragment. The epistemic and non-epistemic circumstances, such as differences in accessible information, or different perspectives, focus, or practical goals etc., of the agent can vary across fragments, which is why these circumstances will be the parameters distinguishing fragments from each other. What exactly this means will become clear throughout this dissertation. For the first qualitative part of the dissertation, justification will be modelled to be a non-monotonic justification relation that holds between pieces of evidence and the propositions they justify, which will be indexed to fragments. Fragments, here, will be modelled by using total plausibility pre-orders over possible worlds. Later on, within a quantitative account of fragmented justification, there will be a variety of ways to spell out the concept of justification. Here, fragments will be modelled by using (subjective) probability functions or credence functions. After an introductory chapter, there will be a chapter focusing on an overview over relevant literature that is being discussed or worked with in this dissertation. Then, chapter 3 presents a skeleton of a qualitative (non-probabilistic) account of fragmented justification and shows how this account is able to provide a novel and attractive solution to the preface paradox (Makinson, 1965), one of the central paradoxes of epistemology. Then, chapter 4 develops a more full fledged account of qualitative fragmented justification, using this time not the preface paradox but a scenario from climate science as a motivating example for the account. Here, a scientific expert about climate science seems to be justified in believing an overall inconsistent set of propositions. The chapter explains how the toolkit of orthodox epistemology is limited in dealing with situations like this, and argues that fragmentation is a more appropriate strategy than existing approaches to deal with (seemingly rational) inconsistent doxastic states. The chapter provides a formalisation of this example, as well as a number of justifying reasons for why fragmentation can be rational. It, also, develops some precautionary norms of fragmentation. It concludes by explaining how fragmented justification relates to belief, and argues that fragmented justification provides a novel rationality argument for accounts of fragmented belief, and, furthermore, how one should go about de-fragmenting in light of practical rationality. Chapters 5 and 6 turn to a probabilistic or quantitative account of fragmented justification -- fragmented Bayesianism. Chapter 5 presents a very general version of fragmented Bayesianism, much more general than the existing accounts of fragmented Bayesianism that already exist. The basic idea is that the doxastic state of an agent will be represented not by one credence function, but by a set of credence functions, corresponding to different fragments. The focal point of this chapter is a list of parameters, such as background knowledge, algebra, confirmation measure, or ur-prior, that are part of the orthodox Bayesianism machinery, which, each, can be chosen to be either fragmented or global within a fragmented account of Bayesianism. Depending on which and how many of those parameters are chosen to be kept global or to render fragmented, the resulting account of fragmented Bayesianism will be more or less severely fragmented. The chapter provides motivations and arguments for why Bayesianism in general, and each of the parameters in particular, deserve fragmented versions. It also provides a brief comparison between the qualitative account of fragmented justification developed in the previous chapters with fragmented Bayesianism. Lastly, chapter 6 compares fragmented Bayesianism with another framework, imprecise Bayesianism. Imprecise Bayesianism aims to model agents who do not have enough information to form precise credences. Fragmentation, on the other hand, aims to represent agents with fragmented doxastic states, where fragments are individuated via differences in the agent’s epistemic and non-epistemic circumstances. The formal models underlying these two accounts are strikingly similar. In both frameworks, the agent's doxastic state is represented not via one but via many different credence functions. The literature on fragmentation and imprecision do not comment on the striking similarity between these two different accounts. This is why chapter 6 is devoted to presenting and comparing their similarities and differences. The dissertation closes with a conclusion and some remarks on potential avenues for further work.

This paper presents a new solution to the preface paradox (Makinson, 1965). It makes use of a fragmented account of justification, and thus, unlike existing solutions, employs the central concept involved in the paradox – justification. First, I present the original paradox, and recall a prominent family of solutions to it – solutions that make use of fragmentation accounts of belief. I argue that, while fragmented belief solutions go in the right direction, what does the work in this paradox is not belief but justification. I then present my account of fragmented justification. I show how it solves the preface paradox in a novel way and I provide reasons for why it is a good solution. The upshot of my account of fragmented justification is that the justification relations that hold between pieces of evidence and the propositions they support are fragment-relative.

In this thesis, I present neighbourhood semantics as a framework for modelling evidence-based belief, as first presented in Van Benthem and Pacuit (2011). I argue why this is a well-suited framework for modelling this particular doxastic attitude, by postulating a number of desiderata that a good formal representation of evidence and evidence-based belief should satisfy, and by showing that neighbourhood semantics outperforms its competitors on almost all of these desiderata. I, further- more, explain and accommodate its shortcomings (i.e. the desideratum it does not fully satisfy) by making a qualification on the framework - I incorporate degrees of trustworthiness of the evidence a belief is based on into the models, to render the belief a justified belief, since for a belief to be justified, it being based on evidence is not sufficient, the evidence needs to, in addition, be trustworthy, i.e. come from a reliable source. After presenting the adapted semantics, I develop an epistemic hierarchy of different types of beliefs, and, I, furthermore, explain how the notion of trustworthiness of evidence I introduce differs from Van Benthem and Pacuit (2011)’s idea of reliable evidence. Lastly, I present and expand on Van Benthem and Pacuit (2011)’s suggestion for group notions of evidence and belief. I present their original notions of group evi- dence and formulate corresponding notions of group belief. In addition, I formulate clauses of group evidence and belief incorporating the notion of trustworthiness, and finally, present group notions of evidence and belief corresponding to the most prominent aggregation rules from the judgment aggregation literature.

In my dissertation, I argue that groups can be justified in believing a contradiction. I present and motivate a Lockean account of justified individual belief and, on this basis, argue for a mechanism for forming justified group belief which is tailored to peer groups of scientific experts about complex systems such as the climate. Because of the special epistemic circumstances due to, both, the type of groups I consider, as well as their areas of research, justified contradictions on the group level are possible, and sometimes even epistemically beneficial. I conclude by arguing that this result is non-standard, because present accounts of justified group belief do not allow for justified contradictions.

The aim of this thesis is to present three puzzles about epistemic modals which arise when adopting an orthodox propositionalist framework for modelling their semantics and pragmatics. I present a stragegy for solving these problems which I call “non-propositionalist”. This strategy consists in constructing an alternative, semi-dynamic framework for epistemic modals which is called “Domain semantics”, as well as a non-orthodox pragmatic picture. Because the non-propositionalist strategy runs into related problems, I present a more general non-propositionalist solution. It uses a fully dynamic semantics while keeping the pragmatic picture of the non-propositionalist relatively unchanged. This general strategy manages to solve the original puzzles while avoiding to run into related problems.

Imprecise Bayesianism, as opposed to orthodox Bayesianism, aims to model agents who do not have enough information to form precise credences. Under this account, an agents' doxastic state is allowed to be made up of not one but many different credence functions, to model imprecision in the assignment of credences to propositions. Fragmentation, on the other hand, aims to represent agents with fragmented doxastic states, where fragments are individuated via differences in the agent's epistemic and non-epistemic circumstances, such as differences in accessible information, or different practical goals and decisions. The formal model underlying this account is identical to imprecise Bayesianism: the agent's doxastic state is represented via not one but many different credence functions. The literature on fragmentation and imprecision do not comment on the striking similarity between these two different accounts. This paper is devoted to presenting and comparing their similarities and differences. The upshot will be that the main difference lies in, first, the interpretation of the formalisms, as well as how these two accounts deal with questions of practical rationality.

This article presents a fragmented account of Bayesianism in its’ most general form. First, it recalls the central theses of orthodox Bayesianism, and provides an overview over the notions of justification that can be for- mulated within orthodox Bayesianism. It presents a very general account of fragmented Bayesianism, and then provides a list of parameters along which the account can be made moderately or very fragmented, depending on which and how many of those parameters one chooses to relativize to fragments. It, furthermore, presents an epistemic risk argument in order to justify fragmen- tation. I first provide a very general version of it, and the provide epistemic risk arguments for the fragmentation of each of the parameters. Lastly, the paper concludes by a brief comparison between the account of fragmented Bayesianism developed here and the qualitative account of fragmented justification developed earlier.

This article presents a fragmented account of justification. The account is motivated via an example from climate science where, within the IPCC assessment report for climate change, an overall inconsistent set of justified beliefs seems to be asserted. The main reason why this happens will be the extreme uncertainty involved in learning about complex systems such as the climate. The article explains how the toolkit of traditional epistemology is limited in and not suited for dealing with situations like this, and argues that fragmentation is a more appropriate strategy than existing approaches to deal with (seemingly rational) inconsistent doxastic states. After presenting the example, the article gives a formalisation of this example. The upshot of this formalisation is that justifications are represented via non-monotonic fragment-indexed justification relations that hold between propositions that need justification, and evidence. The article explains why fragmentation is rationally warranted here, and develops an explicit set of precautionary norms that should guide the fragmentation process. It, furthermore, argues that the account of fragmented justification developed here is compatible with fragmented belief, and how this account of fragmented justification can serve as a novel argument for the rationality of fragmented belief accounts developed in the literature. And lastly, the article explains how sometimes, even though fragmentation might be epistemically rational, de-fragmentation will become necessary in light of practical rationality. It shows how de-fragmentation can work for epistemically rational fragmented agents, which is a novel discussion within the literature on fragmentation.