Hostile epistemology is the study of how environmental features exploit our cognitive vulnerabilities. I am particularly interested in those vulnerabilities arise from the basic character of our epistemic lives. We are finite beings with limited cognitive resources, perpetually forced to reasoning a rush. I focus on two sources of unavoidable vulnerability. First, we need to use cognitive shortcuts and heuristics to manage our limited time and attention. But hostile forces can always game the gap between the heuristic and the ideal. (...) Second, we need to trust others, and trust makes us vulnerable. In particular, we need to trust experts whose expertise lies beyond our individual ability to vet. So we must rely on various imperfect signals and proxies, like institutional affiliation, to manage our trust. We are forced to use these imperfect and exploitable forms of reasoning, in order to cope with a cognitively overwhelming world. I present the hostile epistemology framework as a counterbalance to vice epistemology, which tends to locate responsibility for bad beliefs in the character of the believer. I draw two conclusions from my analysis. First: suppose we accept that that only communities and institutions can know about the world. This deeply social approach doesn’t solve the problem of expert identification. The individual still faces a problem: given their incomplete understanding, how should the individual pick which group to trust? Second: we might have hoped to get some kind of static picture of the right principles of reasoning — but hostile epistemology presents a different picture. We have to use risky heuristics and shortcuts, which hostile forces can game and exploit. We can respond by shifting to a different set of heuristics, but this simply shifts the location of our vulnerabilities. These new heuristics are also exploitable. There is no end to this, only a constant struggle to adapt. Limited beings in a hostile epistemic environment are locked in an unending epistemic arms race. (shrink)

It is well-known that racism is encoded into the social practices and institutions of medicine. Less well-known is that racism is encoded into the material artifacts of medicine. We argue that many medical devices are not merely biased, but materialize oppression. An oppressive device exhibits a harmful bias that reflects and perpetuates unjust power relations. Using pulse oximeters and spirometers as case studies, we show how medical devices can materialize oppression along various axes of social difference, including race, gender, class, (...) and ability. Our account uses political philosophy and cognitive science to give a theoretical basis for understanding materialized oppression, explaining how artifacts encode and carry oppressive ideas from the past to the present and future. Oppressive medical devices present a moral aggregation problem. To remedy this problem, we suggest redundantly layered solutions that are coordinated to disrupt reciprocal causal connections between the attitudes, practices, and artifacts of oppressive systems. (shrink)

Often machine learning programs inherit social patterns reflected in their training data without any directed effort by programmers to include such biases. Computer scientists call this algorithmic bias. This paper explores the relationship between machine bias and human cognitive bias. In it, I argue similarities between algorithmic and cognitive biases indicate a disconcerting sense in which sources of bias emerge out of seemingly innocuous patterns of information processing. The emergent nature of this bias obscures the existence of the bias itself, (...) making it difficult to identify, mitigate, or evaluate using standard resources in epistemology and ethics. I demonstrate these points in the case of mitigation techniques by presenting what I call ‘the Proxy Problem’. One reason biases resist revision is that they rely on proxy attributes, seemingly innocuous attributes that correlate with socially-sensitive attributes, serving as proxies for the socially-sensitive attributes themselves. I argue that in both human and algorithmic domains, this problem presents a common dilemma for mitigation: attempts to discourage reliance on proxy attributes risk a tradeoff with judgement accuracy. This problem, I contend, admits of no purely algorithmic solution. (shrink)

What does it take to be prejudiced against a particular group? And is prejudice always epistemically problematic, or are there epistemically innocent forms of prejudice? In this paper, I argue that certain important forms of prejudice can be wholly constituted by the differential accessibility of certain pieces of information. These accessibility relations constitute a salience structure. A subject is prejudiced against a particular group when their salience structure is unduly organised around that category. This is significant because it reveals that (...) prejudice does not require the presence of any explicit cognitive or emotive attitude, nor need it manifest in behaviour: it can be solely constituted by the organisation of information, where that information may be accurate and well-founded. Nonetheless, by giving an account of ‘undue organisation’ in epistemic terms, I show that this account is compatible with an understanding of prejudice as a negatively valenced epistemic category. (shrink)

In this paper I explore three challenges to the morality of belief. First, whether we have the necessary control over our beliefs to be held responsible for them, i.e., the challenge of doxastic involuntarism. Second, the question of whether belief is really the attitude that we care about in the cases used to motivate the morality of belief. Third, whether attitudes weaker than belief, such as credence, can wrong, I then end by turning to how answers to the previous challenges (...) suggest a way of extending the morality of belief to encompass a way of thinking of the moral mind more generally. (shrink)

As inductive decision-making procedures, the inferences made by machine learning programs are subject to underdetermination by evidence and bear inductive risk. One strategy for overcoming these challenges is guided by a presumption in philosophy of science that inductive inferences can and should be value-free. Applied to machine learning programs, the strategy assumes that the influence of values is restricted to data and decision outcomes, thereby omitting internal value-laden design choice points. In this paper, I apply arguments from feminist philosophy of (...) science to machine learning programs to make the case that the resources required to respond to these inductive challenges render critical aspects of their design constitutively value-laden. I demonstrate these points specifically in the case of recidivism algorithms, arguing that contemporary debates concerning fairness in criminal justice risk-assessment programs are best understood as iterations of traditional arguments from inductive risk and demarcation, and thereby establish the value-laden nature of automated decision-making programs. Finally, in light of these points, I address opportunities for relocating the value-free ideal in machine learning and the limitations that accompany them. (shrink)

“Implicit bias” is a term of art referring to relatively unconscious and relatively automatic features of prejudiced judgment and social behavior. While psychologists in the field of “implicit social cognition” study “implicit attitudes” toward consumer products, self-esteem, food, alcohol, political values, and more, the most striking and well-known research has focused on implicit attitudes toward members of socially stigmatized groups, such as African-Americans, women, and the LGBTQ community.[1] For example, imagine Frank, who explicitly believes that women and men are equally (...) suited for careers outside the home. Despite his explicitly egalitarian belief, Frank might nevertheless implicitly associate women with the home, and this implicit association might lead him to behave in any number of biased ways, from trusting feedback from female co-workers less to hiring equally qualified men over women. Psychological research on implicit bias is relatively recent (§1), but a host of metaphysical (§2), epistemological (§3), and ethical questions (§4) about implicit bias are pressing.[2]. (shrink)

An influential objection to the epistemic power of the imagination holds that it is uninformative. You cannot get more out of the imagination than you put into it, and therefore learning from the imagination is impossible. This paper argues, against this view, that the imagination is robustly informative. Moreover, it defends a novel account of how the imagination informs, according to which the imagination is informative in virtue of its analog representational format. The core idea is that analog representations represent (...) relations ‘for free,’ and this explains how the imagination can contain more information than is put into it. This account makes important contributions to both philosophy of mind, by showing how the imagination can generate new content that is not represented by a subject’s antecedent mental states, and epistemology, by showing how the imagination can generate new justification that is not conferred by a subject’s antecedent evidence. (shrink)

Morality bears on what we should forget. Some aspects of our identity are meant to be forgotten and there is a distinctive harm that accompanies the permanence of some content about us, content that prompts a duty to forget. To make the case that forgetting is an integral part of our moral duties to others, the paper proceeds as follows. In §1, I make the case that forgetting is morally evaluable and I survey three kinds of forgetting: no-trace forgetting, archival (...) forgetting, and siloing. In §2, I turn to how we practice these forms of forgetting in our everyday lives and the goods these practices facilitate by drawing on examples ranging from the expunging of juvenile arrest records to the right to privacy. In §3, I turn to how my account can help us both recognize and address a heretofore neglected source of harm caused by technology and big data. In §4, I end by addressing the concern that we lack control over forgetting and thus can't be required to forget. I argue this challenge can be answered, but there’s a harder challenge that can’t. Forgetting is under threat. To address this challenge and preserve forgetting, we must change the world. (shrink)

A semantics of pictorial representation should provide an account of how pictorial signs are associated with the contents they express. Unlike the familiar semantics of spoken languages, this problem has a distinctively spatial cast for depiction. Pictures themselves are two-dimensional artifacts, and their contents take the form of pictorial spaces, perspectival arrangements of objects and properties in three dimensions. A basic challenge is to explain how pictures are associated with the particular pictorial spaces they express. Inspiration here comes from recent (...) proposals that analyze depiction in terms of geometrical projection. In this essay, I will argue that, for a central class of pictures, the projection-based theory of depiction provides the best explanation for how pictures express pictorial spaces, while rival perceptual and resemblance theories fall short. Since the composition of pictorial space is itself the basis for all other aspects of pictorial content, the proposal provides a natural foundation for further pictorial semantics. (shrink)

Hate speech and misinformation are rife. How to respond? Counterspeech proposals say: with more and better speech. This paper considers the treatment of counterspeech in Maxime Lepoutre’s Democratic Speech In Divided Times. Lepoutre provides a nuanced defence of counterspeech. Some counterspeech, he grants, is flawed. But, he says: counterspeech can be debugged. Once we understand why counterspeech fails – when fail it does – we can engineer more effective counterspeech strategies. Lepoutre argues that the failures of counterspeech can be theorised (...) using the ideology of salience. Negative counterspeech fails because it reinforces the salience of the very ideas or associations that it contests. His solution? Positive counterspeech – a form of counterspeech which avoids the salience trap. I argue that the salience paradigm is ill-suited to theorise the failures of counterspeech. I suggest some alternatives. Further, I show that these alternative paradigms make importantly different practical recommendations – recommendations concerning how we ought to engineer our counterspeech – from those issued by the salience paradigm. (shrink)

The concept of bias is pervasive in both popular discourse and empirical theorizing within philosophy, cognitive science, and artificial intelligence. This widespread application threatens to render the concept too heterogeneous and unwieldy for systematic investigation. This article explores recent philosophical literature attempting to identify a single theoretical category—termed ‘bias’—that could be unified across different contexts. To achieve this aim, the article provides a comprehensive review of theories of bias that are significant in the fields of philosophy of mind, cognitive science, (...) machine learning, and epistemology. It focuses on key examples such as perceptual bias, implicit bias, explicit bias, and algorithmic bias, scrutinizing their similarities and differences. Although these explorations may not conclusively establish the existence of a natural theoretical kind, pursuing the possibility offers valuable insights into how bias is conceptualized and deployed across diverse domains, thus deepening our understanding of its complexities across a wide range of cognitive and computational processes. (shrink)

In this paper, I sketch a general framework for theorizing about bias and bias attributions. According to the account, paradigmatic cases of bias involve systematic departures from genuine norms. I attempt to show that the account illuminates a number of important psychological phenomena, including: the fact that accusations of bias frequently inspire not only denials but also countercharges of bias (“you only think that I'm biased because you're biased!”); the fact that we tend to see ourselves as less biased than (...) our peers (the so‐called ‘bias blind spot’); and the fact that we tend to see people who share our views as less biased than people who don't. I explore the circumstances in which we're committed to believing that those who disagree with us are not only mistaken but also biased simply because they disagree with us in the way that they do. In addition, the account also sheds light on another notorious and well‐documented psychological phenomenon: the fact that introspection is an unreliable way of detecting our biases. On the account that I offer, the unreliability of introspection for this purpose isn't a contingent fact that depends on the finer details of human psychology but rather holds of necessity: even God could not have made us creatures who reliably detect our own biases by way of introspection. (shrink)

This paper explores the complex nature of social biases, arguing for a functional framework that recognizes their unity and diversity. The functional approach posits that all biases share a common functional role in overcoming underdetermination. This framework, I argue, provides a comprehensive understanding of how all psychological biases, including social biases, are unified. I then turn to the question of disunity, demonstrating how psychological social biases differ systematically in the mental states and processes that constitute them. These differences indicate that (...) biases at various levels of the cognitive architecture require distinct treatment along at least two dimensions: epistemic evaluation and mitigation strategies. By examining social biases through this dual lens of unity and diversity, we can more effectively identify when and how to intervene on problematic biases. Ultimately, this approach provides a nuanced understanding of the nature of social bias, offering practical guidance for addressing existing biases and proactively managing emerging biases in both human and artificial minds. (shrink)

One of the core insights from Eleanor Rosch’s work on categorization is that human categorization isn’t arbitrary. Instead, two psychological principles constrain possible systems of classification for all human cultures. According to these principles, the task of a category system is to provide maximum information with the least cognitive effort, and the perceived world provides us with structured rather than arbitrary features. In this paper, I show that Rosch's insights give us important resources for making progress on the 'feasibility question' (...) in conceptual engineering: the question of how we can implement conceptual engineering projects in ways that are practically feasible. Specifically, I show that one overlooked upshot of Rosch's work is that naming practices play an extremely important role in the construction of perceived similarities within and dissimilarities between categories, and, correspondingly, the dissemination of social stereotypes that serve as markers between different categories that are otherwise similar. Thus, naming practices will be a crucial constraint for the feasibility of certain ameliorative projects. (shrink)

Recent findings in different areas of psychology and cognitive science have brought the unconscious mind back to center stage. However, the unconscious mind worry remains: What renders unconscious phenomena mental? I suggest a new strategy for answering this question, which rests on the idea that categorizing unconscious phenomena as “mental” should be scientifically useful relative to the explanatory research goals. I argue that this is the case if by categorizing an unconscious phenomenon as “mental” one picks out explanatorily relevant similarities (...) to a corresponding paradigmatically mental conscious phenomenon. Explanatory relevance is spelled out in terms of mechanistic norms. (shrink)

Implicit attitudes are mental states posited by psychologists to explain behaviors including implicit racial and gender bias. In this paper I investigate the belief view of the implicit attitudes, on which implicit attitudes are a kind of implicit belief. In particular, I focus on why implicit attitudes, if they are beliefs, are often resistant to updating in light of new evidence. I argue that extant versions of the belief view do not give a satisfactory account of this phenomenon. This is (...) because proponents of the belief view have focused on overly narrow explanations of recalcitrance in terms of belief storage. Expanding the focus of the belief view to the kinds of irrational and arational transitions between beliefs and other mental states provides compelling (if preliminary) explanations of recalcitrance. (shrink)

Epistemic analyses of stereotyping describe how they lead to misperceptions and misunderstandings of social actors and events. The analyses have tended so far to focus on how people acquire stereotypes and/or how the stereotypes lead to distorted perceptions of the evidence that is available about individuals. In this chapter, I focus instead on how the stereotypes can generate misleading evidence by influencing the policy preferences of people who harbour the biases. My case study is stereotypes that relate to people living (...) in poverty. I show how these stereotypes influence policy choices in ways that generate misleading evidence about people living in poverty. I argue that the stereotypes generate the misleading evidence by supporting policies that restrict the agency of the people in poverty. In generating this misleading evidence, the stereotypes place additional constraints on the epistemic agency of everyone, making it harder for anyone, including those who do and those who do not endorse the stereotypes, to gain true beliefs about people living in poverty. Going forward, I conclude, adequate epistemic analyses of stereotyping ought to be more expansive, acknowledging both the way that stereotypes generate misleading evidence by constraining the agency of those stereotyped, and how we can all thereby be epistemically constrained by the stereotypes harboured by others. (shrink)

The possibilities of unconscious perception and unconscious bias prompt parallel debates about unconscious mental content. This chapter argues that claims within these debates alleging the existence of unconscious content are made fraught by ambiguity and confusion with respect to the two central concepts they involve: consciousness and content. Borrowing conceptual resources from the debate about unconscious perception, the chapter distills the two conceptual puzzles concerning each of these notions and establishes philosophical strategies for their resolution. It then argues that empirical (...) evidence for unconscious bias falls victim to these same puzzles, but that progress can be made by adopting similar philosophical strategies. Throughout, the chapter highlights paths forward in both debates, illustrates how they serve as fruitful domains in which to study the relationship between philosophy and empirical science, and uses their combined study to further understanding of a general theory of unconscious content. (shrink)

We introduce the Bias Network Approach (BNA) as a sociotechnical method for AI developers to identify, map, and relate biases across the AI development process. This approach addresses the limitations of what we call the "isolationist approach to AI bias," a trend in AI literature where biases are seen as separate occurrence linked to specific stages in an AI pipeline. Dealing with these multiple biases can trigger a sense of excessive overload in managing each potential bias individually or promote the (...) adoption of an uncritical approach to understanding the influence of biases in developers’ decision-making. The BNA fosters dialogue and a critical stance among developers, guided by external experts, using graphical representations to depict biased connections. To test the BNA, we conducted a pilot case study on the "waiting list” project, involving a small AI developer team creating a healthcare waiting list NPL model in Chile. The analysis showed promising findings: (i) the BNA aids in visualizing interconnected biases and their impacts, facilitating ethical reflection in a more accessible way; (ii) it promotes transparency in decision-making throughout AI development; and (iii) more focus is necessary on professional biases and material limitations as sources of bias in AI development. (shrink)

In recent years, the epistemic reliability of moral intuitions has been undermined by substantial empirical data reporting the influence of cognitive biases. This paper discusses and elaborates upon a promising strategy in response to the reliability challenge to moral intuitions. The considered argument appeals to the fact that moral intuitions are experienced with different levels of strength and agents accept only strong intuitions, not vulnerable to bias under realistic circumstances. This essay aims to reconstruct this defense from the reliability challenge (...) in its most promising form and to evaluate the plausibility of the argument in light of the available empirical evidence. What emerges from the discussion is that the vindication of moral intuitions fundamentally depends on two distinct premises: first, the hypothesis that agents accept moral intuitions proportionally to their level of confidence, and second, the hypothesis that intuitive confidence is epistemically reliable. While there is consistent evidence for the first hypothesis, there is still no conclusive evidence for the second. (shrink)

What makes discrimination morally bad? In this paper, we discuss the putative badness of a case of consensual discrimination to show that prominent accounts of the badness of discrimination—appealing, inter alia, to harm, disrespect and inequality—fail to provide a satisfactory answer to this question. In view of this, we present a more promising account.

What attitude does someone manifesting implicit bias really have? According to the default representationalist picture, implicit bias involves having conflicting attitudes (explicit versus implicit) with respect to the topic at hand. In opposition to this orthodoxy, dispositionalists argue that attitudes should be understood as higher-level dispositional features of the person as a whole. Following this metaphysical view, the discordance characteristic of implicit bias shows that someone’s attitude regarding the topic at hand is not-fully-manifested or ‘in-between’. However, so far few representationalists (...) have been convinced by dispositionalist arguments, largely because dispositionalism cannot provide explanations in terms of underlying processes. We argue that if dispositionalism wants to be a genuine contender, it should make clear what it has to offer in terms of understanding of implicit bias. As a concrete proposal, we combine dispositionalist metaphysics with the idea that our normative practices of attitude ascription partly determine what it means to have an attitude. We show that such regulative dispositionalism can account for two prominent normative features of implicit bias. We conclude by suggesting that in order to engage in a meaningful debate with representationalism, dispositionalists might have to put the question ‘what counts as a good explanation?’ back on the table. (shrink)

Feminist philosophers have long worried that intersectionality undermines the viability of the concept and category of woman, thereby undermining feminist theory and politics. Some have responded to this problem by abandoning intersectionality; others have attempted to find some suitably inclusive way of reconceptualizing woman. I provide a novel solution that focuses on conceptualizing oppression in light of intersectionality, rather than trying to provide an account of what it is to be a woman. By enabling us to understand feminism as responding (...) to gender oppression, this account shows that intersectionality does not conceptually undermine and fragment feminism. Feminism should be intersectional. (shrink)

Large language models (LLMs) such as OpenAI’s ChatGPT reflect, and can potentially perpetuate, social biases in language use. Conceptual engineering aims to revise our concepts to eliminate such bias. We show how machine learning and conceptual engineering can be fruitfully brought together to offer new insights to both conceptual engineers and LLM designers. Specifically, we suggest that LLMs can be used to detect and expose bias in the prototypes associated with concepts, and that LLM de-biasing can serve conceptual engineering projects (...) that aim to revise such conceptual prototypes. At present, these de-biasing techniques primarily involve approaches requiring bespoke interventions based on choices of the algorithm’s designers. Thus, conceptual engineering through de-biasing will include making choices about what kind of normative training an LLM should receive, especially with respect to different notions of bias. This offers a new perspective on what conceptual engineering involves and how it can be implemented. And our conceptual engineering approach also offers insight, to those engaged in LLM de-biasing, into the normative distinctions that are needed for that work. (shrink)

In order to uncover the inner workings of our capacities, we look to ‘effects’. Most of us have the capacity to distinguish between spoken ‘ba’ and ‘fa’ sounds. One thought is that this is achieved through aural sensitivities that detect changes in vibration picked up by the eardrum. But the McGurk Effect suggests that there is more to the story. Without changing the incoming vibrations, sound experience can be modulated by showing a video of a mouth making a ‘ba’ sound (...) or a ‘fa’ sound with a consistent sound overlaid. We learn that our overall auditory experiences are at least in part determined by visual cues in addition to what’s first picked up by our eardrums. The McGurk Effect gives us a hint into the inner workings of audition and helps us better understand the capacity to discriminate sounds of a certain sort. -/- In the present paper, the focus is on emotional capacities and a well known effect – recalcitrance. Recalcitrant emotions, such as fearing the dog even though one knows that the dog is harmless or being angry with one’s partner even when one realises it was only in a dream that the partner was nasty, have played the role of effect in much theorising about emotions. But in my view, we’ve stayed a bit too close to home, aiming to fit the effect into a paradigm – the representationalist paradigm – that isn’t fit for purpose. I will use this criticism as a launching off point to introduce a different way of thinking about emotions that is better suited to making sense of recalcitrance. I will argue that emotions are transitions between representational states rather than being representational states themselves. The view is better suited to make sense of recalcitrance and, at the end of the paper, I will offer reasons for thinking that main points that speak in favour of a representationalist approach to emotion can be recaptured or explained away by the transitions view. (shrink)

In this chapter, we identify and present predominant debates at the intersection of ethics and imagination. We begin by examining issues on whether our imagination can be constrained by ethical considerations, such as the moral evaluation of imagination, the potential for morality’s constraining our imaginative abilities, and the possibility of moral norms’ governing our imaginings. Then, we present accounts that posit imagination’s integral role in cultivating ethical lives, both through engagements with narrative artworks and in reality. Our final topic of (...) consideration focuses on the possibility of imagination constituting or constructing new ethical or political frameworks. (shrink)