Colin Klein

YouTube has been implicated in the transformation of users into extremists and conspiracy theorists. The alleged mechanism for this radicalizing process is YouTube’s recommender system, which is optimized to amplify and promote clips that users are likely to watch through to the end. YouTube optimizes for watch-through for economic reasons: people who watch a video through to the end are likely to then watch the next recommended video as well, which means that more advertisements can be served to them. This is a seemingly innocuous design choice, but it has a troubling side-effect. Critics of YouTube have alleged that the recommender system tends to recommend extremist content and conspiracy theories, as such videos are especially likely to capture and keep users’ attention. To date, the problem of radicalization via the YouTube recommender system has been a matter of speculation. The current study represents the first systematic, pre-registered attempt to establish whether and to what extent the recommender system tends to promote such content. We begin by contextualizing our study in the framework of technological seduction. Next, we explain our methodology. After that, we present our results, which are consistent with the radicalization hypothesis. Finally, we discuss our findings, as well as directions for future research and recommendations for users, industry, and policy-makers.

Generative artificial intelligence systems based on transformers, including both text-generators like GPT-4 and image generators like DALL-E 3, have recently entered the popular consciousness. These tools, while impressive, are liable to reproduce, exacerbate, and reinforce extant human social biases, such as gender and racial biases. In this paper, we systematically review the extent to which DALL-E Mini suffers from this problem. In line with the Model Card published alongside DALL-E Mini by its creators, we find that the images it produces tend to represent dozens of different occupations as populated either solely by men (e.g., pilot, builder, plumber) or solely by women (e.g., hairdresser, receptionist, dietitian). In addition, the images DALL-E Mini produces tend to represent most occupations as populated primarily or solely by White people (e.g., farmer, painter, prison officer, software engineer) and very few by non-White people (e.g., pastor, rapper). These findings suggest that exciting new AI technologies should be critically scrutinized and perhaps regulated before they are unleashed on society.

Computers are used to make decisions in an increasing number of domains. There is widespread agreement that some of these uses are ethically problematic. Far less clear is where ethical problems arise, and what might be done about them. This paper expands and defends the Ethical Gravity Thesis: ethical problems that arise at higher levels of analysis of an automated decision-making system are inherited by lower levels of analysis. Particular instantiations of systems can add new problems, but not ameliorate more general ones. We defend this thesis by adapting Marr’s famous 1982 framework for understanding information-processing systems. We show how this framework allows one to situate ethical problems at the appropriate level of abstraction, which in turn can be used to target appropriate interventions.

The epistemic position of an agent often depends on their position in a larger network of other agents who provide them with information. In general, agents are better off if they have diverse and independent sources. Sullivan et al. [19] developed a method for quantitatively characterizing the epistemic position of individuals in a network that takes into account both diversity and independence; and presented a proof-of-concept, closed-source implementation on a small graph derived from Twitter data [19]. This paper reports on an open-source reimplementation of their algorithm in Python, optimized to be usable on much larger networks. In addition to the algorithm and package, we also show the ability to scale up our package to large synthetic social network graph profiling, and finally demonstrate its utility in analyzing real-world empirical evidence of ‘echo chambers’ on online social media, as well as evidence of interdisciplinary diversity in an academic communications network.

There is not as much resistance to COVID-19 mitigation as there seems, but there are structural features that make resistance seem worse than it is. Here we describe two ways that the problem seeming to be worse than it is can make it worse. First, visible hesitation to implement COVID-19 responses signals to the wider society that mitigation measures may not succeed, which undermines people’s conditional willingness to join in on those efforts. Second, our evaluations of others’ willingness to implement these measures are informed by our attempts to mind-read them. Yet attempts to mind-read groups often mislead us, because groups invariably act from diverse motives whereas mind-reading works best when identifying relatively stable and consistent motivations. This means that a small minority of people refusing to implement measures can have an outsized prominence that prompts mind-reading to diagnose widespread hesitation. These two factors form a feedback loop with each other: we see some people’s hesitation, which prompts us to mind-read other people as being more uncertain about the responses than they actually are, which undermines our confidence in the responses, which in turn encourages others to mind-read this hesitation, which further undermines confidence.

The social media platform Twitter platform has played a crucial role in the Black Lives Matter (BLM) movement. The immediate, flexible nature of tweets plays a crucial role both in spreading information about the movement’s aims and in organizing individual protests. Twitter has also played an important role in the right-wing reaction to BLM, providing a means to reframe and recontextualize activists’ claims in a more sinister light. The ability to bring about social change depends on the balance of these two forces, and in particular which side can capture and maintain sustained attention. The present study examines 2 years worth of tweets about BLM (about 118 million in total). Timeseries analysis reveals that activists are better at mobilizing rapid attention, whereas right-wing accounts show a pattern of moderate but more sustained activity driven by reaction to political opponents. Topic modeling reveals differences in how different political groups talk about BLM. Most notably, the murder of George Floyd appears to have solidified a right-wing counter-framing of protests as arising from dangerous “terrorist” actors. The study thus sheds light on the complex network and rhetorical effects that drive the struggle for online attention to the BLM movement.

The medium‐independence of computational descriptions has shaped common conceptions of computational explanation. So long as our goal is to explain how a system successfully carries out its computations, then we only need to describe the abstract series of operations that achieve the desired input–output mapping, however they may be implemented. It is argued that this abstract conception of computational explanation cannot be applied to so‐called real‐time computing systems, in which meeting temporal deadlines imposed by the systems with which a device interfaces are constitutive of the computing tasks that a device performs. Instead, real‐time computing reveals the need for alternative conceptions of computational explanation, as well as computational implementation, that eschew medium‐independence.

In this paper we analyses the work of the first century Chinese philosopher Wang Chong as in part grappling with epistemology of testimony. Often portrayed as a curmudgeonly skeptic, Wang Chong actually best seen as a demanding piecemeal non-reductionist, which is to say he believed that testimony was a basic source of evidence unless subject to a defeater (non-reductionism), but also that we should evaluate testimony on a claim-by-claim basis (piecemeal) rather than accepting a whole source on the strength of its reputation or authority, and finally he thought one has a strong duty to search for defeaters (demanding). We defend each of these claims as well as the fruitfulness of applying such anachronistic terminology to an ancient thinker. We end with a discussion of how Wang Chong's epistemology relates to his rhetorical method and then consider some problems he faces because of his epistemic commitments.

Functional neuroimaging (NI) technologies like Positron Emission Tomography and functional Magnetic Resonance Imaging (fMRI) have revolutionized neuroscience, and provide crucial tools to link cognitive psychology and traditional neuroscientific models. A growing discipline of 'neurophilosophy' brings fMRI evidence to bear on traditional philosophical issues such as weakness of will, moral psychology, rational choice, social interaction, free will, and consciousness. NI has also attracted critical attention from psychologists and from philosophers of science. I review debates over the evidential status of fMRI, including the differences between brain scans and ordinary images, the legitimacy of forward inference and reverse inference, and deductive versus probabilistic accounts of NI evidence. I conclude with a discussion of fMRI as exploratory rather than confirmatory evidence, linking this debate to the growing literature on cognitive ontology.

The dual-track theory of moral reasoning has received considerable attention due to the neuroimaging work of Greene et al. Greene et al. claimed that certain kinds of moral dilemmas activated brain regions specific to emotional responses, while others activated areas specific to cognition. This appears to indicate a dissociation between different types of moral reasoning. I re-evaluate these claims of specificity in light of subsequent empirical work. I argue that none of the cortical areas identified by Greene et al. are functionally specific: each is active in a wide variety of both cognitive and emotional tasks. I further argue that distinct activation across conditions is not strong evidence for dissociation. This undermines support for the dual-track hypothesis. I further argue that moral decision-making appears to activate a common network that underlies self-projection: the ability to imagine oneself from a variety of viewpoints in a variety of situations. I argue that the utilization of self-projection indicates a continuity between moral decision-making and other kinds of complex social deliberation. This may have normative consequences, but teasing them out will require careful attention to both empirical and philosophical concerns

fMRI promises to uncover the functional structure of the brain. I argue, however, that pictures of ‘brain activity' associated with fMRI experiments are poor evidence for functional claims. These neuroimages present the results of null hypothesis significance tests performed on fMRI data. Significance tests alone cannot provide evidence about the functional structure of causally dense systems, including the brain. Instead, neuroimages should be seen as indicating regions where further data analysis is warranted. This additional analysis rarely involves simple significance testing, and so justified skepticism about neuroimages does not provide reason for skepticism about fMRI more generally

Consciousness supervenes on activity; computation supervenes on structure. Because of this, some argue, conscious states cannot supervene on computational ones. If true, this would present serious difficulties for computationalist analyses of consciousness (or, indeed, of any domain with properties that supervene on actual activity). I argue that the computationalist can avoid the Superfluous Structure Problem (SSP) by moving to a dispositional theory of implementation. On a dispositional theory, the activity of computation depends entirely on changes in the intrinsic properties of implementing material. As extraneous structure is not required for computation, a system can implement a program running on some but not all possible inputs. Dispositional computationalism thus permits episodes of computational activity that correspond to potential episodes of conscious awareness. The SSP cannot be motivated against this account, and so computationalism may be preserved

Protests and counter-protests seek to draw and direct attention and concern with confronting images and slogans. In recent years, as protests and counter-protests have partially migrated to the digital space, such images and slogans have also gone online. Two main ways in which these images and slogans are translated to the online space is through the use of emoji and hashtags. Despite sustained academic interest in online protests, hashtag activism and the use of emoji across social media platforms, little is known about the specific functional role that emoji and hashtags play in online social movements. In an effort to fill this gap, the current paper studies both hashtags and emoji in the context of the Twitter discourse around the Black Lives Matter movement.

Pain asymbolics feel pain, but act as if they are indifferent to it. Nikola Grahek argues that such patients present a clear counterexample to motivationalism about pain. I argue that Grahek has mischaracterized pain asymbolia. Properly understood, asymbolics have lost a general capacity to care about their bodily integrity. Asymbolics’ indifference to pain thus does not show something about the intrinsic nature of pain ; it shows something about the relationship between pains and subjects, and how that relationship might break down. I explore the consequences of such a view for both motivationalism and the categorization of pain asymbolia as a syndrome, arguing for a close link between asymbolia and various forms of depersonalization.

Unlike the overall framework of Ernest Nagel's work on reduction, his theory of intertheoretic connection still has life in it. It handles aptly cases where reduction requires complex representation of a target domain. Abandoning his formulation as too liberal was a mistake. Arguments that it is too liberal at best touch only Nagel's deductivist theory of explanation, not his condition of connectability. Taking this condition seriously gives a powerful view of reduction, but one which requires us to index explanatory power to sciences as they are formulated at particular times. While we may thereby reduce more than philosophers have supposed, we must abandon hope (as Nagel did) of saying anything useful about reductionism

Multiply realizable properties are those whose realizers are physically diverse. It is often argued that theories which contain them are ipso facto irreducible. These arguments assume that physical explanations are restricted to the most specific descriptions possible of physical entities. This assumption is descriptively false, and philosophically unmotivated. I argue that it is a holdover from the late positivist axiomatic view of theories. A semantic view of theories, by contrast, correctly allows scientific explanations to be couched in the most perspicuous, powerful language available. On a semantic view, traditional notions of multiple realizability are thus very hard to motivate. At best, one must abandon either the idea that multiple realizability is an interesting scientific notion, or else admit that multiply realizable properties do not automatically block scientific reductions

Hypocrites invite moral opprobrium, and charges of hypocrisy are a significant and widespread feature of our moral lives. Yet it remains unclear what hypocrites have in common, or what is distinctively bad about them. We propose that hypocrites are persons who have undermined their claim to moral authority. Since this self-undermining can occur in a number of ways, our account construes hypocrisy as multiply realizable. As we explain, a person’s moral authority refers to a kind of standing that they occupy within a particular moral community. This status is both socially important and normatively precarious. Hence, moral agents are right to be vigilant when it comes to hypocrisy, and are often justified in their outrage when they detect it. We further argue that our view can preserve what is attractive in rival accounts, while avoiding their associated problems.

Some vegetative state patients show fMRI responses similar to those of healthy controls when instructed to perform mental imagery tasks. Many authors have argued that this provides evidence that such patients are in fact conscious, as response to commands requires intentional agency. I argue for an alternative reading, on which responsive patients have a deficit similar to that seen in severe forms of akinetic mutism. Akinetic mutism is marked by the inability to form and maintain intentions to act. Responsive patients are likely still conscious. However, the route to this conclusion does not support attributions of intentional agency. I argue that aspects of consciousness, rather than broad diagnostic categories, are the more appropriate target of empirical investigation. Investigating aspects of consciousness provides a better method for investigating profound disorders of consciousness. 1 Introduction2 Responses in the Vegetative State2.1 The imaging evidence2.2 The need for models3 Responsive Patients and Akinetic Mutism3.1 Akinetic mutism3.2 Akinetic mutism as a deficit in intentions3.3 Arguments for the link3.4 Interim conclusion4 Other Models4.1 A deficit in ability?4.2 Modular intentions?5 Consciousness and Method5.1 Are they conscious?5.2 Aspects versus levels.

Multiply realizable kinds are scientifically problematic, for it appears that we should not expect discoveries about them to hold of other members of that kind. As such, it looks like MR kinds should have no place in the ontology of the special sciences. Many resist this conclusion, however, because we lack a positive account of the role that certain realization-unrestricted terms play in special science explanations. I argue that many such terms actually pick out idealizing models. Idealizing explanation has many of the features normally associated with explanation by MR kinds. As idealized models are usually mere possibilia, such explanations do not run afoul of the metaphysical problems that plague MR kinds.

In What the Body Commands, Colin Klein proposes and defends a novel theory of pain. Klein argues that pains are imperative; they are sensations with a content, and that content is a command to protect the injured part of the body. He terms this view "imperativism about pain," and argues that imperativism can account for two puzzling features of pain: its strong motivating power and its uninformative nature. Klein argues that the biological purpose of pain is homeostatic; like hunger and thirst, pain helps solve a challenge to bodily integrity. It does so by motivating you to act in ways that help the body recover. If you obey pain's command, you get better. He develops his account to handle a variety of pain phenomena and applies it to solve a number of historically puzzling cases. Klein's intent is to defend the imperativist view in a pure form -- without requiring pain to represent facts about the world.Klein presents a model of imperative content showing that intrinsically motivating sensations are best understood as imperatives, and argues that pain belongs to this class. He considers the distinction between pain and suffering; explains how pain motivates; addresses variations among pains; and offers an imperativist account of maladaptive pains, pains that don't appear to hurt, masochism, and why pain feels bad.

The so-called “dark room problem” makes vivd the challenges that purely predictive models face in accounting for motivation. I argue that the problem is a serious one. Proposals for solving the dark room problem via predictive coding architectures are either empirically inadequate or computationally intractable. The Free Energy principle might avoid the problem, but only at the cost of setting itself up as a highly idealized model, which is then literally false to the world. I draw at least one optimistic conclusion, however. Real-world, real-time systems may embody motivational states in a variety of ways consistent with idealized principles like FEP, including ways that are intuitively embodied and extended. This may allow predictive coding theorists to reconcile their account with embodied principles, even if it ultimately undermines loftier ambitions.

Since its introduction, multivariate pattern analysis, or ‘neural decoding’, has transformed the field of cognitive neuroscience. Underlying its influence is a crucial inference, which we call the decoder’s dictum: if information can be decoded from patterns of neural activity, then this provides strong evidence about what information those patterns represent. Although the dictum is a widely held and well-motivated principle in decoding research, it has received scant philosophical attention. We critically evaluate the dictum, arguing that it is false: decodability is a poor guide for revealing the content of neural representations. However, we also suggest how the dictum can be improved on, in order to better justify inferences about neural representation using MVPA.

The interpretation of functional imaging experiments is complicated by the pluripotency of brain regions. As there is a many-to-one mapping between cognitive functions and their neural substrates, region-based analyses of imaging data provide only weak support for cognitive theories. Price and Friston argue that we need a ‘cognitive ontology’ that abstractly categorizes the function of regions. I argue that abstract characterizations are unlikely to be cognitively interesting. I argue instead that we should attribute functions to regions in a context-sensitive manner. I review recent meta-analyses that approach fMRI data in this light and argue that they have revisionary potential