Files in this item



application/pdfIan_Harmon.pdf (1MB)
(no description provided)PDF


Title:Epistemological implications of representational pluralism
Author(s):Harmon, Ian
Director of Research:Cummins, Robert
Doctoral Committee Chair(s):Cummins, Robert
Doctoral Committee Member(s):Waskan, Jonathan; Roth, Martin; Melnick, Arthur
Department / Program:Philosophy
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):non-propositional knowledge
practical rationality
Mental representation
Abstract:In this dissertation I argue that the framework under which epistemology operates should be broadened to account for developments in cognitive science that indicate that a good deal of cognition and reasoning involves the use of non-linguistic representations. In chapter 1, I argue that, although epistemology is the theory of knowledge, epistemologists generally operate as though their field is simply the theory of propositional knowledge. Epistemologists generally assume that knowledge is a certain type of belief relation to a true proposition. However, cognitive science indicates that many of our mental representations are not belief-like at all, and thereby, not belief relations to propositions. Rather, the mind employs representations that take the form of images, scale models, activation patterns, and so on. I call this claim representational pluralism. If some of these non-linguistic representations are constitutive of knowledge, as I argue that they are in later chapters, then this requires a substantial revision of the traditional epistemological framework. I proceed to introduce some potential consequences of departing from the propositional knowledge tradition in epistemology. These consequences pertain most directly to two issues, namely, philosophical methodology and our understanding of normative standards of rationality. My discussion of methodological issues begins with the introduction of what I call the analysis problem. The analysis problem is the problem of developing a conceptual analysis of propositional knowledge and justified belief. I argue that this problem emerged from concerns with skeptical regress arguments and the Gettier problem. The traditional methodology for analyzing these concepts, and hence, for addressing the analysis problem, has consisted in determining the truth conditions for attributions of knowledge and justified beliefs. To determine these truth conditions, epistemologists develop thought experiments designed to elicit our semantic intuitions regarding the use of “knowledge,” “justified belief,” and their respective cognates. After enough intuitions are elicited, we formulate a theory of knowledge or justification. While this methodology is problematic in its own right, as I argue in chapters 2 and 3, it has also lead to the embrace of an understanding of rationality that is limited and narrow in its application. Rationality amounts to epistemic constraint satisfaction and epistemic constraints provide criteria that allow us to assess an agent’s behavior for good or correct performance. Good, or correct, performance must be understood relative to a specific goal or problem. Since epistemology assumes that knowledge is a certain belief relation to a true proposition, the goal or problem of concern in epistemology is that of forming true beliefs. So, when epistemologists think about rationality, they think of behaving in a way that is conducive to forming true beliefs. In other words, the standards of rationality that epistemologists use to evaluate agents for rationality apply only to agents who exhibit behavior that is aimed at forming true beliefs. So, one reason that the traditional understanding of rationality in epistemology is problematic is because it is inapplicable to agents who have other goals, both epistemic and practical. The second problem I discuss for epistemology’s traditional understanding of rationality is that the standards of rationality are formulated in abstract conditions that idealize away from various practical constraints to which we are constantly subject. It is important to keep in mind that this second problem is, in some sense orthogonal to the first, although it arrives, at least at times, from the standard epistemological methodology. Nevertheless, even if the traditional epistemological framework is correct in holding that all knowledge is belief-like, epistemology’s traditional standards of rationality are inapplicable to any actual individuals because actual individuals are finite creatures with fixed cognitive architectures. Since epistemology’s traditional standards of rationality are formulated in idealized conditions that abstract away from these limitations, they should be understood as guiding the behavior of agents who exist in those sorts of environments. But agents who exists in those sorts of environments will be subject to more demanding standards than actual agents. Actual agents, of course, cannot meet these standards, and, since ought implies can, cannot be held to them. In short, chapter 1 argues that we need to depart from the epistemological status quo in order to accommodate the possibility of non-belief-like, that is, non-propositional knowledge. The dissertation proceeds to explore some of the consequences of this departure, specifically the consequences for philosophical methodology and our understanding of rationality in epistemology. In chapter 2 I discuss knowledge-how, a type of knowledge that many have argued is non-propositional. In the first part of the chapter I discuss two intellectualist positions, that is, positions that hold that know-how is propositional. Stanley and Williamson argue that propositional knowledge is both necessary and sufficient for know-how, while Bengson and Moffett argue that propositional knowledge is not sufficient, but is necessary, for know-how. Stanley and Williamson’s position is that knowing how to X amounts to knowing, under a practical mode of presentation, that some way is a way for one to X. Against their position, I argue that if practical modes of presentation do the work that is required of them, then positing them amounts to granting that non-propositional knowledge is necessary for know-how. According to Stanley and Williamson, practical modes of presentation explain the connection between know-how, dispositions, and actions. However, these dispositions are presumably those that enable performance, and in order to have these dispositions, I argue that we need to practice the activity in question. Through this practice, we acquire the content that gives rise to the relevant dispositions. Following representational pluralism, it is highly implausible that this content is always linguistic or propositional. Hence, positing practical modes of presentation amounts to positing non-propositional content. Bengson and Moffett concede that there is a non-propositional element that is necessary for know-how. Specifically, they argue that in order to know how to X, one has to stand in a non-propositional knowledge relation to a way of X-ing. However, they also argue that knowing how to X requires propositional knowledge because one can stand in a non-propositional relation to a way of X-ing without knowing that it is a way of X-ing, and thereby fail to know how to X. As I argue, their position entails that many clear-cut cases of knowing how are not cases of knowing how because the relevant agents do not know that the way in which they X is the way in which they X. Taken together, my discussion of Stanley and Williamson and Bengson and Moffett shows that propositional knowledge is neither necessary nor sufficient for know-how. The second half of chapter 2 discusses the origins of the contemporary know-how debate going back to Gilbert Ryle. I argue that, since Ryle’s time, two debates have been taking place in the know-how literature: one regarding the semantic analysis of know-how ascriptions and another regarding how we ought to explain various skills or abilities. The first debate is concerned with truth conditions for sentences such as, “Hannah knows how to ride a bicycle,” while the latter debate is concerned with determining whether skills are the result of applying a “theory” or stored propositional knowledge, or are the result of the processing of non-linguistic information. Unfortunately, these debates have been entangled in the literature. For instance, some philosophers, including Stanley and Williamson, Bengson and Moffett, and Ryle, have made inferences about how we ought to explain skills or abilities on the basis of semantic analyses of know-how ascriptions or appeals to ordinary language. Both of these strategies assume that language gives us the correct truth conditions for sentences that use mental terminology and that scientific accounts of the mind are beholden to language. In other words, these strategies amount to what Martin Roth and Robert Cummins call epistemological poaching. As an alternative approach that disentangles these debates, I argue that the representational pluralist thesis needs to be taken seriously. The use of epistemological poaching tacitly assumes that the thesis is false and that cognition is structured around a language of thought that has a similar structure to natural language. Taking representational pluralism seriously can allow epistemology to develop a broader, yet more specialized, framework that bridges the longstanding gap between the field and empirical approaches to understanding knowledge. In chapter 3 I argue that epistemology’s failure to take representational pluralism seriously has skewed the field’s understanding of normative standards of rationality. I discuss two ways in which epistemology’s normative standards of rationality are limited. First, they apply only to agents with purely linguistic or belief-like cognitive systems. Second, they apply only to cognitive systems that are capable of meeting them, due to what I call the ought-can principle. But before discussing these limits on epistemology’s normative standards of rationality, I consider one way in which a proponent of traditional epistemology might try to argue that representational pluralism does not require substantial revision of epistemology’s framework. I label this type of argument the doxastification strategy. More specifically the doxastification strategy constitutes an attempt to argue that apparent cases of non-propositional knowledge can be accommodated within epistemology’s propositional framework. Before discussing the doxastification strategy in depth, I present prima facie evidence for what I call epistemological pluralism, the thesis that there are many types of knowledge, many of which are non-propositional. The first type of prima facie evidence for epistemological pluralism comes from representational pluralism. If representational pluralism is true, there appears to be no reason to hold that only linguistic representations can be constitutive of knowledge. The second type of prima facie evidence comes from semantics. We often make knowledge attributions in which it appears the thing known is not a proposition. For instance, we say things like, “Jones knows how the New York subway system is laid out,” or “Jones knows what “God Only Knows” sounds like. In the first case, the thing known appears to be, not a proposition, but the layout of a subway system. In the second case the thing known appears to be, not a proposition, but the sound of a song. Taken together, we can take semantic data to give us an indication as to what sorts of things qualify as objects of knowledge. Meanwhile, the psychological evidence provides us with insight regarding how these objects are mentally represented. The layout of a subway system is naturally represented by a map, while the tune to a song is naturally represented in auditory memory or in a score. Hence, it appears, at least prima facie, that we have knowledge that is, in part, comprised of non-linguistic mental representations. One might, however, employ the doxastification strategy to argue that we can understand these cases of apparent non-propositional knowledge in terms of propositional knowledge. When one knows, for instance, how the New York subway system is laid out, what one knows is a proposition, namely, the proposition that the New York subway system is laid out like this, where “this” refers to a map of the subway system. However, the doxastification strategy divorces the content that does the genuine evidential work from the content of what is known. Certainly knowing how the New York subway system is laid out allows one to know that the subway system is laid out like this (again, where “this” refers to a map of the subway system). But the problem is that one can know this proposition without having any idea how the New York subway system is laid out. One can be reliably informed that the map in question accurately represents the subway system’s layout without ever taking a look at the map and thereby know that the subway system is laid out like that. But in order to know how the subway system is laid out, one needs access to an actual representation of the system, that is a map (mental or otherwise). After discussing doxastification, I argue the problems with epistemology’s standards of rationality can be resolved by reconciling the field with the relevant findings in cognitive science, that is, findings that support representational pluralism. . I argue that the current standards of rationality in epistemology apply neither to individuals nor to collectives or institutional cognitive systems. These standards cannot be applied to the individual because doing so violates the principle that ought implies can. In other words, these standards require that individuals exceed their capacities. These standards cannot be applied to institutions, such as the institution of science, because they are not equipped to evaluate the use of non-linguistic representations that is ubiquitous in scientific reasoning. Next, I discuss several problems with traditional methodology in epistemology, and motivate an alternative approach. The first problem is that it is not clear what the target of a conceptual analysis of knowledge is. If one thinks we need an analysis that is correct in all possible worlds, then it is not clear what could possibly constrain such a project. Though appealing to intuition is a standard approach, there does not appear to be any reason to think that intuitions are equipped to provide any evidential support for a theory of knowledge. I argue that, instead of pursuing an analysis of knowledge in the traditional manner, we should ask what role knowledge plays in the various domains in which it is employed. In chapter 4, I examine the role that knowledge plays in two domains: everyday life and the institution of science. In everyday contexts, I argue that knowledge plays a warrant-granting role for action. This way of thinking about knowledge has drawn some attention in the epistemology literature from Keith DeRose, Jason Stanley, John Hawthorne, and Jeremy Fantl and Matthew McGrath, amongst others. However, I argue that none of these “pragmatic encroachment” approaches draw the correct connection between knowledge and practical affairs. In particular, many of these accounts are designed to be supplements to more traditional, independent accounts of knowledge. Rather, I suggest, that in order to take pragmatic encroachment seriously, knowledge needs to be understood as that which plays the role of making an action rational under realistic conditions in which time and memory are limited. That which makes an action rational is often different than that which makes a belief rational. Forming beliefs is generally low risk, that is, there is little cost that comes with being wrong. The cost of being wrong comes into play only when we act on beliefs. But when we are simply concerned with the formation of beliefs, and not how we ought to act on beliefs, the risks we undertake are generally minimal. Because of these differences, the formation of beliefs is subject to different standards of rationality than is acting on beliefs. When we are genuinely concerned with determining how to act, standards of rationality must be sensitive to limits of time, memory, information, and other resources. Hence, I argue, drawing on work by Gigerenzer and Goldstein, that what makes an action rational in everyday contexts is not a proposition that is known in the traditional philosophical sense, but rather the use of an effective algorithm or set of heuristics. If this is correct, then knowledge, as understood in traditional epistemology, cannot fill the role that knowledge plays in everyday contexts. Though propositions can serve as inputs into a decision-making procedure, it is effective use of the procedure itself that makes an action rational. In other words, it is an algorithm or decision procedure that fills the role that knowledge plays in everyday contexts. It is true, of course, that in everyday circumstances, knowledge is often used or discussed in a way that is much closer to the way it is understood in standard epistemology. It does not seem out place in everyday contexts to say, for instance, that I know that Jefferson City is the capital of Missouri. I take this to suggest that wee need to understand knowledge in a pluralistic way. While I do think knowledge is often used in everyday contexts to justify or criticize action, it is unlikely that this captures all uses. Ultimately, the issue of what role knowledge plays in commonsense is an empirical question, and so it would be most desirable to accumulate a set of linguistic data to give us a clearer sense of the different roles the concept plays in ordinary usage. In scientific contexts, I argue that, because science aims to provide us with an understanding of the world, the role that knowledge plays in science is an explanatory role. However, it is possible to have an explanation for a phenomenon that is not correct or accurate. Such an explanation demonstrates, not how actually a phenomenon occurs, but rather how possibly or how plausibly the phenomenon occurs. Since science aims to provide theories that are not only explanatory, but also correct or accurate, the role of knowledge in science is that of an accurate explanatory role. I begin by noting that Hempel’s Deductive-Nomological account of explanation and some subsequent accounts assume explanations are sentence-like in structure. But more recently, many philosophers of science have taken a mechanistic approach to explanation. Machamer, Darden, and Craver, for instance, hold that representations of mechanisms for phenomena explain those phenomena. They note that we use diagrams to represent features of mechanisms, and these diagrams allow us to more easily apprehend the phenomena than linguistic descriptions. While some philosophers of science, such as Carl Craver and J.D. Trout, deny that explanations have to render their target happenings intelligible, Waskan e al. present compelling empirical evidence that this view is not shared by practicing scientists. Rather, it appears that both scientists and laypersons have a concept of explanation in which intelligibility plays a central role. This supports Machamer, Darden, and Craver’s claim that representations of mechanisms render their target happenings intelligible. There is still a good deal of work to be done to determine the psychological nature of intelligibility or understanding. However, Stephen Grimm suggests that understanding a phenomenon requires having a grasp of the relevant scientific principles and the ability to apply these principles to specific cases. If knowing how entails possessing content that gives rise to dispositions that enable us to perform various abilities, as I argue in chapter 2, then Grimm provides some reason for thinking that know-how is a central component to understanding. So if knowledge plays the role of explaining the world, and explanations must render the world intelligible, then it appears that the role of knowledge in science is filled by that which enables us to have a correct or accurate understanding of the world. If this is correct, then the role that knowledge plays in science is clearly different than the role it plays in everyday contexts. In everyday contexts, knowledge provides warrant for action. In science, knowledge explains the world by rendering it intelligible. This difference should be unsurprising given that individuals have different goals, resources, and limitations than practicing scientists, or even scientific institutions. However, one common feature that knowledge in science and everyday life appear to have in common is that they are both related to abilities or skills. In some everyday contexts, knowledge takes the form of a decision procedure of algorithm. Having propositional information that can serve as an input to an algorithm alone does not provide sufficient warrant or assurance that a certain course of action is rational. To have a sufficient degree of warranty, one also needs an effective procedure for processing or utilizing that information. To put it in somewhat more standard terms, it would seem that knowing how to put the information at one’s disposal to practical use is essential for making a course of action practically rational. In scientific contexts, knowledge plays the role of enabling us to understand the world, and if Grimm is correct, then understanding consists, at least in part, of a kind of knowledge-how. There is one final complication that I discuss at the end of the chapter. The complication is that science is generally an institutional or collective endeavor. At the individual level, it appears that the cognitive state that plays the role of knowledge in science will be that of a state of understanding, which Grimm suggests entails having certain skills or abilities. But since science is generally a collective enterprise, following the present proposal will require developing an account of understanding, ability, and skill that makes sense when applied to collective cognitive systems. At present, I am not sure what such an account will look like. However, there are at least two obvious possibilities. One possibility is that at the collective or institutional level, understanding, skill, and ability are to be understood in roughly the same manner as they are understood as the individual level. For example, we might understand the utterance, “Science understands how planes fly” as serving as shorthand for the claim that a sufficient number of individual scientists, or a sufficient number of members of a subset of scientists, understand how planes fly. A second possibility would be that in order for science to understand a phenomenon, it could be the case that different scientists, or collections of scientists, understand different components of that phenomenon. Perhaps none of the scientists, or groups of scientists, understands the phenomenon in full, but if we sum together the individual components that they do separately understand, we will arrive at a full understanding of the phenomenon under consideration. If the suggestions offered in this section are correct, then they constitute further support for the claim that epistemology stands in need of revision. The traditional epistemological framework is not equipped to evaluate knowledge under the role it plays in science and this role is distinct from the role that it plays in common sense. Hence, we have further need for epistemological specialization and fragmentation. In chapter 5, I present summaries of the preceding chapters. I then conclude that epistemology, in order to be the theory of knowledge, rather than the theory of propositional knowledge, must become more pluralistic. This is not to say that we need to abandon studies of propositional knowledge. But rather, epistemology as a field should become more fragmented, specialized, and connected to scientific accounts of the mind and cognition.
Issue Date:2014-09-16
Rights Information:Copyright 2014 Ian Harmon
Date Available in IDEALS:2014-09-16
Date Deposited:2014-08

This item appears in the following Collection(s)

Item Statistics