Search

Empirical justification is the justification for our beliefs that is grounded in experience. In this dissertation, I defend three aspects of empirical justification that stem from Descartes' epistemological framework: foundationalism, individualism, and the epistemic privilege of experiences. Chapter 1 is an overview of the dissertation. Chapter 2 argues that contemporary theorists have overstated the case against the epistemic privilege of experiences, though I point out that there are compelling reasons for thinking this connection ought to be understood in a fallibilistic manner. Chapter 3 discusses contemporary theories of evidence and proposes a Cartesian-inspired account of evidence. Chapter 4 turns the typical discussion of skepticism and empirical justification on its head: instead of explaining how ordinary beliefs are justified in the face of skeptical threats, I aim to explain the structure of justified beliefs in skeptical threats given the falsity of skepticism qua philosophical doctrine. Chapter 5 discusses a plethora of morally loaded and ideological cases that are sometimes invoked in an attempt to cast doubt on individualistic accounts of epistemic justification. After critically analyzing these cases, I advance an individualistic theory that accommodates them.

Most people think that empathy is a valuable trait, that (all things equal) an empathetic person is a good person. Recently this idea has been challenged by psychologists and philosophers who are worried by findings that indicate a general tendency to empathize more reliably with people who are similar to us than those who are not. This strikes a blow to those who would have hoped that empathy is precisely the sort of thing that can bridge difference. In this dissertation, I aim to challenge the implicit notion that empathy’s value comes from its usefulness as a tool. Empathy can be a tool, true, but it is not just that. It is a way of relating to others that is valuable for what it is and, perhaps, what it expresses, in addition to what it does. In Chapter 1, I aim to provide an answer to the titular question: “How Should We Think About Empathy?” The first answer is that we should think about it in terms of what I call the AAA Model, which is to say, we should understand empathy as having an attention-attribution-acquisition structure. When we empathize, we first attend to affective indicators, then make an emotion attribution on the basis of those indicators, before finally acquiring an isomorphic emotional state in virtue of making the attribution. This structure allows for plenty of flexibility, and I therefore describe this way of conceptualizing empathy as a generic approach, a framework within which other theorists could be narrower in their conceptions of empathy according to their aims and preferences. Crucially, I also argue that we should think of empathy as one of a number of fellow feelings – affective responses that bear variations on an attention-attribution-acquisition structure. In Chapter 2, “Reasons for Empathy,” I introduce a way of thinking about empathy non-instrumentally through the concept of fittingness. The recent literature in philosophy of emotion has converged on a separation between fitting and non-fitting (i.e., value-based, instrument

I defend the surprising claim that scientists often do—and always should—reason about complex adaptive systems as if they were bound by normative standards that exist independently of (a) the scientists’ own normative commitments and (b) the actual features of the complex adaptive systems (including their normative commitments, if they’re capable of such). Following David Enoch, I refer to such standards as “robustly normative”. Chapters 1 and 2 show that molecular biologists and neuroethologists reason as though certain kinds of functions and mechanisms, respectively, are robustly normative. Chapters 3 and 4 show that (a) psychologists who study logical and moral reasoning, respectively, do not reason as though rationality were robustly normative and (b) this creates deep problems that cannot be solved unless they do reason as though rationality were robustly normative. Through my dissertation, I suggest an even stronger, metaphysical claim: that scientists always should reason about complex adaptive systems as if they were bound by robust norms because they are, in fact, bound by robust norms. Chapter 1 develops a template for this argument: it explains how normativity emerges from the causal structure of the photoreceptor protein rhodopsin and argues that this emergent structure licenses the kinds of explanatory roles that normativity plays in the complexity sciences. Future work will modify this template for the other explanatory roles that I attribute to norms in Chapters 2–4.

In the philosophy of science, there have been a number of attempts to characterize high-level phenomena as patterns in lower-level phenomena. In computer science, there has been a great deal of effort devoted to recognizing patterns and to representing those patterns using formal models. This dissertation engages with the philosophical literature on patterns from a perspective informed by relevant work in computer science. My aim is not to correct philosophers' errors or misconceptions by appeal to results in computer science---though that may occasionally be appropriate. Rather, my aim is to bolster the philosophical literature by introducing helpful conceptual tools drawn from fields like machine learning, computational modeling, pattern recognition, and algorithmic information theory. As I argue, these conceptual tools allow us to place pattern-based accounts of non-fundamental ontology (esp. Dennett, 1991) on a firmer theoretical foundation. More concretely, I propose a novel, rigorous, and scientifically-informed criterion for the reality of patterns in the physical world. I also propose a more general approach to understanding non-fundamental ontology using ideas from machine learning, such as features and feature selection. This approach to non-fundamental ontology, I argue, will allow us to better understand scientific modeling, inter-level relations, scientific reduction, and other important issues in the philosophy of science.

There is something that it is like to be us as we go about our waking and dreaming lives. There is something that it is like to see red, to feel a pressure, or to taste hot coffee. Each of our experiences has a phenomenal character. Our conscious experiences also represent the world around us; they have representational or intentional contents. An experience as of a moderate pressure against one’s shin represents a force being exerted against that part of one’s body. If there is a force of that degree at that location, then the experience is accurate; otherwise, it is inaccurate. In addition to representing the world around us, we also represent the phenomenal characters of our experiences by thinking about them. We can prefer some experiences to others, for instance. This dissertation concerns the role of phenomenal consciousness in the representational mind. It explores the connections between what our experiences are like, what those experiences represent, and how we think about those experiences. In particular, it explores various structural connections between the phenomenal characters and intentional contents of our mental states. In ‘The Price of Twin Earth’, I consider the extent to which the representational contents of our perceptual experiences are reflected in their phenomenal characters. I do so by taking the debate between liberals and conservatives about the contents of perceptual experience as a case study and showing how to generalize those results. Liberals claim that our perceptual experiences (a) can represent specific, familiar individuals and kinds as such, and (b) that these representations are reflected in the phenomenal characters of our perceptual experiences: those of our perceptual experiences that represent their object as being one familiar individual or kind differ in their phenomenal character from those that represent their object as being a different individual or kind or that do not represent their object as being a familiar indivi

Imagine that as you are sitting down to eat a sandwich, you hear someone issue a command: “Stop that this instant!” Perplexed, you look up to see a stranger glaring at you, and come to believe that they told you to stop eating. But in the midst of your ongoing interaction with this person, your thought about the stranger doesn’t seem to represent them as either a mere object or agent. Instead, it seems natural to think of them in a second-person way, as a “you” or an agent with whom I am interacting. I argue that during face-to-face interactions with other agents, some of our thoughts represent these agents in an irreducibly second-person way. My dissertation defends an account of these irreducibly second-person thoughts. Chapter 2 introduces second-person thought. I argue that irreducibly second-person thought employs an irreducibly second-person mode of presentation, and that this mode of presentation should itself be understood in terms of its function: recognizing its object as an agent with whom I am engaged in interaction. This recognition-based account is superior to competing accounts that model the second-person mode of presentation in terms of paradigmatic interactions, like communicative interactions (e.g., linguistic exchanges which employ the second-person pronoun ‘you’), or cooperative interactions. Chapters 3 and 4 defend a basic argument for the irreducibility of second-person thought. Among others, John Perry and David Lewis argue that irreducibly first-person thought—variously called essentially indexical thought or de se thought—plays an essential role in the explanation of action and yields a distinctive form of self-knowledge. While prominent defenses of the irreducibility of such thought extend only to self-directed thought, chapter 3 develops an analogous argument for thought about other agents, contending that irreducibly second-person thought plays an analogous role in the explanation of interaction and yields a distinctive form of knowledge

Though the performance of many applications is dominated by memory behavior, our ability to describe, capture, compare, and recreate that behavior is quite limited. This inability underlies much of the complexity in the field of performance analysis: it is fundamentally difficult to relate benchmarks and applications or use realistic workloads to guide system design and procurement. A concise, observable, and machine-independent characterization of memory behavior is needed. This dissertation presents the Chameleon framework, an integrated solution to three classic problems in the field of memory performance analysis: reference locality modeling, accurate synthetic address trace generation, and the creation of synthetic benchmark proxies for applications. The framework includes software tools to capture a concise, machine-independent memory signature from any application and produce synthetic memory address traces that mimic that signature. It also includes the Chameleon benchmark, a fully tunable synthetic executable whose memory behavior can be dictated by these signatures. By simultaneously modeling both spatial and temporal locality, Chameleon produces uniquely accurate, general- purpose synthetic traces. Results demonstrate that the cache hit rates generated by each synthetic trace are nearly identical to those of the application it targets on dozens of memory hierarchies representing many of today's commercial offerings. This work focuses on the unique challenges of high-performance computing (HPC) where workload selection, benchmarking, system procurement, performance prediction, and application analysis present important challenges. The Chameleon framework can aid in each scenario by providing a concise representation of the memory requirements of full-scale applications that can be tractably captured and accurately mimicked

Though the performance of many applications is dominated by memory behavior, our ability to describe, capture, compare, and recreate that behavior is quite limited. This inability underlies much of the complexity in the field of performance analysis: it is fundamentally difficult to relate benchmarks and applications or use realistic workloads to guide system design and procurement. A concise, observable, and machine-independent characterization of memory behavior is needed. This dissertation presents the Chameleon framework, an integrated solution to three classic problems in the field of memory performance analysis: reference locality modeling, accurate synthetic address trace generation, and the creation of synthetic benchmark proxies for applications. The framework includes software tools to capture a concise, machine-independent memory signature from any application and produce synthetic memory address traces that mimic that signature. It also includes the Chameleon benchmark, a fully tunable synthetic executable whose memory behavior can be dictated by these signatures. By simultaneously modeling both spatial and temporal locality, Chameleon produces uniquely accurate, general- purpose synthetic traces. Results demonstrate that the cache hit rates generated by each synthetic trace are nearly identical to those of the application it targets on dozens of memory hierarchies representing many of today's commercial offerings. This work focuses on the unique challenges of high-performance computing (HPC) where workload selection, benchmarking, system procurement, performance prediction, and application analysis present important challenges. The Chameleon framework can aid in each scenario by providing a concise representation of the memory requirements of full-scale applications that can be tractably captured and accurately mimicked

Several benchmarks for measuring memory performance of HPC systems along dimensions of spatial and temporal memory locality have recently been proposed. However, little is understood about the relationships of these benchmarks to real applications and to each other. In this paper, we propose a methodology for producing architecture-neutral characterizations of the spatial and temporal locality exhibited by the memory access patterns of applications. We demonstrate that the results track intuitive notions of spatial and temporal locality on several synthetic and application benchmarks. We employ the methodology to analyze the memory performance components of the HPC Challenge Benchmarks, the Apex-MAP benchmark, and their relationships to each other and other benchmarks and applications. We show that this analysis can be used to both increase understanding of the benchmarks and enhance their usefulness by mapping them, along with applications, to a 2-D space along axes of spatial and temporal locality.

Several benchmarks for measuring memory performance of HPC systems along dimensions of spatial and temporal memory locality have recently been proposed. However, little is understood about the relationships of these benchmarks to real applications and to each other. In this paper, we propose a methodology for producing architecture-neutral characterizations of the spatial and temporal locality exhibited by the memory access patterns of applications. We demonstrate that the results track intuitive notions of spatial and temporal locality on several synthetic and application benchmarks. We employ the methodology to analyze the memory performance components of the HPC Challenge Benchmarks, the Apex-MAP benchmark, and their relationships to each other and other benchmarks and applications. We show that this analysis can be used to both increase understanding of the benchmarks and enhance their usefulness by mapping them, along with applications, to a 2-D space along axes of spatial and temporal locality.

This thesis consists of two parts. In the first part, we focus on modeling and forecasting arrival rates to a US commercial bank's call center. In today's economy, call centers have become the primary point of contact between customers and businesses. Accurate prediction of the call arrival rate is therefore indispensable for call center practitioners to staff their call center efficiently and cost effectively. We propose a, multiplicative model for modeling and forecasting within-day arrival rates. Markov chain Monte Carlo sampling methods are used to estimate both latent states and model parameters. One-day-ahead density forecasts for the rates and counts are provided. The calibration of these predictive distributions is evaluated through probability integral transforms. Furthermore, we provide one-day-ahead forecasts comparisons with classical statistical models. Our predictions show significant improvements of up to 25% over these standards. A sequential Monte Carlo algorithm is also proposed for sequential estimation and forecasts of the model parameters and rates. Finally, the effect of parameter uncertainty is analyzed in forecasting both the rates and call volumes. The second part focuses on yield curve modeling which has received extensive attention amongst academics and practitioners alike. Until recently though, very little effort was focused on forecasting the yield curve, which plays an important role in risk management, derivative pricing and portfolio management. Diebold and Li(2005) address this issue by fitting a dynamic version of the Nelson-Siegel Curve and compare the out-of-sample performance of this model with several existing models studied in the literature. We extend this work by considering more elaborate dynamics for the three latent factors. Markov chain Monte Carlo sampling methods are used to estimate both latent states and model parameters. In addition, one-step, six-step and twelve-step ahead forecast densities of the yields are provi

This paper tells the story leading to ICANN's selection of seven new Internet top level domains in November 2000. In implementing proposals to expand the name space, ICANN adopted an approach far different from Jon Postel's lightweight proposals. ICANN staff, in setting the ground rules for considering new gTLDs, emphasized that only a few applicants would be allowed in, and imposed strict threshold requirements. Staff determined that the Board should pick TLDs by looking at all relevant aspects of every proposal, and deciding which ones presented the best overall combination of a variety of incommensurable factors. Aspects of the resulting process were predictable: Anyone familiar with the FCC comparative hearing process for broadcast licenses can attest that this sort of ad hoc comparison is necessarily subjective, lending itself to arbitrariness and biased application. Yet the process had advantages that appealed to ICANN decision-makers. The Board members would be free to take their best shots, in a situationally sensitive manner, at advancing the policies they thought important. The approach allowed ICANN to maintain the greatest degree of control. The end result, though, was a process stunning in its arbitrariness, a bad parody of fact-bound, situationally sensitive (rather than rules-based) decision-making., Comment: v.3; only slightly revised from v.2; originally presented at 29th TPRC Conference, 2001

This study presents an integrated approach using experimental high resolution electron microscopy, computer simulations and theoretical analysis, to provide a thorough atomic level investigation of grain boundaries. Examined here, are specific cases of pure aluminum grain boundaries. Energy calculations for the Sigma3 perfect twin, in a range of linear translational states, show that both the shape and location of the interfacial surface are important considerations.

This study presents an integrated approach using experimental high resolution electron microscopy, computer simulations and theoretical analysis, to provide a thorough atomic level investigation of grain boundaries. Examined here, are specific cases of pure aluminum grain boundaries. Energy calculations for the Sigma3 perfect twin, in a range of linear translational states, show that both the shape and location of the interfacial surface are important considerations.