Sina Fazelpour

AI integration into scientific communities promises accelerated discovery but raises concerns about detrimental homogenization. We develop an NK landscape model to explore these promises and risks. We find that non-personalized AI systems that offer uniform guidance yield benefits only under a narrow conjunction of specific problem structure, practices, and baseline research capabilities, becoming harmful otherwise. We implement two proposed mitigations: randomization and personalization. While randomization's utility remains restricted to decomposable problems, personalization preserves or enhances diversity, enabling benefits across conditions. Overall, our results highlight the importance of shifting perspectives in systemic AI evaluations from tool adoption to institutional adaptation.

Recent philosophical work has explored how the social identity of knowers influences how their contributions are received, assessed, and credited. However, a critical gap remains regarding the role of technology in mediating and enabling communication within today’s epistemic communities. This paper addresses this gap by examining how social media platforms and their recommendation algorithms shape the professional visibility and opportunities of researchers from minority groups. Using agent-based simulations, we investigate this question with respect to components of a widely used recommendation algorithm, and uncover three key patterns: First, these algorithms disproportionately harm the professional visibility of researchers from minority groups, creating systemic patterns of exclusion. Second, within these minority groups, the algorithms result in greater visibility for users who more closely resemble the majority group, incentivizing assimilation at the cost of professional invisibility. Third, even for topics that strongly align with minority identities, content created by minority researchers is less visible to the majority than similar content produced by majority users. Importantly, these patterns emerge, even though individual engagement with professional content is independent of group identity. These findings have significant implications for philosophical discussions on epistemic harms that hinder the participation of minority groups, and for policy proposals aimed at addressing these harms. More broadly, they call for a closer examination of the pervasive, but often neglected role of AI and data-driven technologies in shaping today’s epistemic communities.

Recent studies suggest large language models (LLMs) can generate human-like responses, aligning with human behavior in economic experiments, surveys, and political discourse. This has led many to propose that LLMs can be used as surrogates or simulations for humans in social science research. However, LLMs differ fundamentally from humans, relying on probabilistic patterns, absent the embodied experiences or survival objectives that shape human cognition. We assess the reasoning depth of LLMs using the 11-20 money request game. Nearly all advanced approaches fail to replicate human behavior distributions across many models. The causes of failure are diverse and unpredictable, relating to input language, roles, safeguarding, and more. These results warrant caution in using LLMs as surrogates or for simulating human behavior in research.

Disagreements are widespread across the design, evaluation, and alignment pipelines of artificial intelligence (AI) systems. Yet, standard practices in AI development often obscure or eliminate disagreement, resulting in an engineered homogenization that can be epistemically and ethically harmful, particularly for marginalized groups. In this paper, we characterize this risk, and develop a normative framework to guide practical reasoning about disagreement in the AI lifecycle. Our contributions are two-fold. First, we introduce the notion of perspectival homogenization, characterizing it as a coupled ethical-epistemic risk that arises when an aspect of an AI system's development unjustifiably suppresses disagreement and diversity of perspectives. We argue that perspectival homogenization is best understood as a procedural risk, which calls for targeted interventions throughout the AI development pipeline. Second, we propose a normative framework to guide such interventions, grounded in lines of research that explain why disagreement can be epistemically beneficial, and how its benefits can be realized in practice. We apply this framework to key design questions across three stages of AI development tasks: when disagreement is epistemically valuable; whose perspectives should be included and preserved; how to structure tasks and navigate trade-offs; and how disagreement should be documented and communicated. In doing so, we challenge common assumptions in AI practice, offer a principled foundation for emerging participatory and pluralistic approaches, and identify actionable pathways for future work in AI design and governance.

As artificial intelligence (AI) systems increasingly permeate processes of cultural and epistemic production, there are growing concerns about how their outputs may confine individuals and groups to static or restricted narratives about who or what they could be. In this paper, we advance the discourse surrounding these concerns by making three contributions. First, we introduce the concept of aspirational affordance to describe how culturally shared interpretive resources can shape individual cognition, and in particular exercises practical imagination. We show how this concept can ground productive evaluations of the risks of AI-enabled representations and narratives. Second, we provide three reasons for scrutinizing of AI's influence on aspirational affordances: AI's influence is potentially more potent, but less public than traditional sources; AI's influence is not simply incremental, but ecological, transforming the entire landscape of cultural and epistemic practices that traditionally shaped aspirational affordances; and AI's influence is highly concentrated, with a few corporate-controlled systems mediating a growing portion of aspirational possibilities. Third, to advance such a scrutiny, we introduce the concept of aspirational harm, which, in the context of AI systems, arises when AI-enabled aspirational affordances distort or diminish available interpretive resources in ways that undermine individuals' ability to imagine relevant practical possibilities and alternative futures. Through three case studies, we illustrate how aspirational harms extend the existing discourse on AI-inflicted harms beyond representational and allocative harms, warranting separate attention. Through these conceptual resources and analyses, this paper advances understanding of the psychological and societal stakes of AI's role in shaping individual and collective aspirations.

This paper examines two prominent formal trade-offs in artificial intelligence (AI)---between predictive accuracy and fairness, and between predictive accuracy and interpretability. These trade-offs have become a central focus in normative and regulatory discussions as policymakers seek to understand the value tensions that can arise in the social adoption of AI tools. The prevailing interpretation views these formal trade-offs as directly corresponding to tensions between underlying social values, implying unavoidable conflicts between those social objectives. In this paper, I challenge that prevalent interpretation by introducing a sociotechnical approach to examining the value implications of trade-offs. After providing a set-theoretic understanding of these model trade-offs, I argue that their implications cannot be understood in isolation from the technical, psychological, organizational, and social factors that shape the integration of AI models into decision-making pipelines. Specifically, I identify three key considerations---validity and instrumental relevance, compositionality, and dynamics---for contextualizing and characterizing these implications. These considerations reveal that the relationship between model trade-offs and corresponding values depends on critical choices and assumptions. Crucially, judicious sacrifices in one model property for another can, in fact, promote both sets of corresponding values. The proposed sociotechnical perspective thus shows that we can and should aspire to higher epistemic and ethical possibilities than the prevalent interpretation suggests, while offering practical guidance for achieving those outcomes. Finally, I draw out the broader implications of this perspective for AI design and governance, highlighting the need to broaden normative engagement across the AI lifecycle, develop legal and auditing tools sensitive to sociotechnical considerations, and rethink the vital role and appropriate structure of interdisciplinary collaboration in fostering a responsible AI workforce.

How do social factors affect group learning in diverse populations? Evidence from cognitive science gives us some insight into this question, but is generally limited to showing how social factors play out in small groups over short time periods. To study larger groups and longer time periods, we argue that we can combine evidence about social factors from cognitive science with agent-based models of group learning. In this vein, we demonstrate the usefulness of idealized models of inquiry, in which the assumption of Bayesian agents is used to isolate and explore the impact of social factors. We show that whether a certain social factor is beneficial to the community’s epistemic aims depends on its particular manifestation by focusing on the impacts of homophily – the tendency of individuals to associate with similar others – on group inquiry.

Diversity of social identities can improve the performance of groups through varied cognitive and communicative pathways. Recently, research efforts have focused on identifying when we should expect to see these potential benefits in real-world settings. While most research to date has studied this topic at individual and interpersonal levels, in this paper, we develop an agent-based model to explore how various aspects of homophily, the tendency of individuals to associate with similar others, affects performance at a larger scale. Study 1 examines how two types of homophily---identity-driven and opinion-driven---impact collective performance on a sequential decision-making task via modulating network formation and trust relations. Study 2 considers how the presence of identity-based conformity pressure can affect the findings from the first study. Overall, we find that the effect of homophily on performance is complex, depending on the operative dimensions of similarity, mediating pathways, and the specific outcome of interest. Finally, we discuss the implications of our results for policy interventions aiming to improve group performance.

There has been a surge of recent interest in sociocultural diversity in machine learning research. Currently, however, there is a gap between discussions of measures and benefits of diversity in machine learning, on the one hand, and the broader research on the underlying concepts of diversity and the precise mechanisms of its functional benefits, on the other. This gap is problematic because diversity is not a monolithic concept. Rather, different concepts of diversity are based on distinct rationales that should inform how we measure diversity in a given context. Similarly, the lack of specificity about the precise mechanisms underpinning diversity’s potential benefits can result in uninformative generalities, invalid experimental designs, and illicit interpretations of findings. In this work, we draw on research in philosophy, psychology, and social and organizational sciences to make three contributions: First, we introduce a taxonomy of different diversity concepts from philosophy of science, and explicate the distinct epistemic and political rationales underlying these concepts. Second, we provide an overview of mechanisms by which diversity can benefit group performance. Third, we situate these taxonomies of concepts and mechanisms in the lifecycle of sociotechnical machine learning systems and make a case for their usefulness in fair and accountable machine learning. We do so by illustrating how they clarify the discourse around diversity in the context of machine learning systems, promote the formulation of more precise research questions about diversity’s impact, and provide conceptual tools to further advance research and practice.

The predictive coding (PC) theory of attention identifies attention with the optimization of the precision weighting of prediction error. Here we provide some challenges for this identification. On the one hand, the precision weighting of prediction error is too broad a phenomenon to be identified with attention because such weighting plays a central role in multimodal integration. Cases of crossmodal illusions such as the rubber hand illusion and the McGurk effect involve the differential precision weighting of prediction error, yet attention does not shift as one would predict. On the other hand, the precision weighting of prediction error is too narrow a phenomenon to be identified with attention, because it cannot accommodate the full range of attentional phenomena. We review criticisms that PC cannot account for volitional attention and affect-biased attention, and we propose that it may not be able to account for feature-based and intellectual attention.

Machine learning algorithms are increasingly used to shape high-stake allocations, sparking research efforts to orient algorithm design towards ideals of justice and fairness. In this research on algorithmic fairness, normative theorizing has primarily focused on identification of “ideally fair” target states. In this paper, we argue that this preoccupation with target states in abstraction from the situated dynamics of deployment is misguided. We propose a framework that takes dynamic trajectories as direct objects of moral appraisal, highlighting three respects in which such trajectories can be subject to evaluation in relation to their temporal dynamics, robustness, and representation.

Typically, fair machine learning research focuses on a single decision maker and assumes that the underlying population is stationary. However, many of the critical domains motivating this work are characterized by competitive marketplaces with many decision makers. Realistically, we might expect only a subset of them to adopt any non-compulsory fairness-conscious policy, a situation that political philosophers call partial compliance. This possibility raises important questions: how does partial compliance and the consequent strategic behavior of decision subjects affect the allocation outcomes? If k% of employers were to voluntarily adopt a fairness-promoting intervention, should we expect k% progress (in aggregate) towards the benefits of universal adoption, or will the dynamics of partial compliance wash out the hoped-for benefits? How might adopting a global (versus local) perspective impact the conclusions of an auditor? In this paper, we propose a simple model of an employment market, leveraging simulation as a tool to explore the impact of both interaction effects and incentive effects on outcomes and auditing metrics. Our key findings are that at equilibrium: (1) partial compliance by k% of employers can result in far less than proportional (k%) progress towards the full compliance outcomes; (2) the gap is more severe when fair employers match global (vs local) statistics; (3) choices of local vs global statistics can paint dramatically different pictures of the performance vis-a-vis fairness desiderata of compliant versus non-compliant employers; (4) partial compliance based on local parity measures can induce extreme segregation. Finally, we discuss implications for auditors and insights concerning the design of regulatory frameworks.

We suggest that philosophical accounts of epistemic effects of diversity have given insufficient attention to the relationship between demographic diversity and information elaboration, the process whereby knowledge dispersed in a group is elicited and examined. We propose an analysis of IE that clarifies hypotheses proposed in the empirical literature and their relationship to philosophical accounts of diversity effects. Philosophical accounts have largely overlooked the possibility that demographic diversity may improve group performance by enhancing IE, and sometimes fail to explore the relationship between diversity and IE altogether. We claim these omissions are significant from both a practical and theoretical perspective. Moreover, we explain how the overlooked explanations suggest that epistemic benefits of diversity can depend on epistemically unjust social dynamics.

In this paper we argue that predictive processing (PP) theory cannot account for the phenomenon of affect-biased attention prioritized attention to stimuli that are affectively salient because of their associations with reward or punishment. Specifically, the PP hypothesis that selective attention can be analyzed in terms of the optimization of precision expectations cannot accommodate affect-biased attention; affectively salient stimuli can capture our attention even when precision expectations are low. We review the prospects of three recent attempts to accommodate affect with tools internal to PP theory: Miller and Clark’s (2018) embodied inference; Seth’s (2013) interoceptive inference; and Joffily and Coricelli’s (2013) rate of change of free energy. In each case we argue that the account does not resolve the challenge from affect-biased attention. For this reason, we conclude that prediction error minimization is not sufficient to explain all mental phenomena, contrary to the claim that the PP framework provides a unified theory of all mental phenomena or the brain ‘s cognitive functioning. Nevertheless, we suggest that empirical investigation of the interaction between affective salience and precision expectations should prove helpful in understanding the limits of PP theory, and may provide new directions for the application of a Bayesian perspective to perception.

A concept of diversity is an understanding of what makes a group diverse that may be applicable in a variety of contexts. We distinguish three diversity concepts, show that each can be found in discussions of diversity in science, and explain how they tend to be associated with distinct epistemic and ethical rationales. Yet philosophical literature on diversity among scientists has given little attention to distinct concepts of diversity. This is significant because the unappreciated existence of multiple diversity concepts can generate unclarity about the meaning of “diversity,” lead to problematic inferences from empirical research, and obscure complex ethical-epistemic questions about how to define diversity in specific cases. We illustrate some ethical-epistemic implications of our proposal by reference to an example of deliberative mini-publics on human tissue biobanking.

Data‐driven algorithms are widely used to make or assist decisions in sensitive domains, including healthcare, social services, education, hiring, and criminal justice. In various cases, such algorithms have preserved or even exacerbated biases against vulnerable communities, sparking a vibrant field of research focused on so‐called algorithmic biases. This research includes work on identification, diagnosis, and response to biases in algorithm‐based decision‐making. This paper aims to facilitate the application of philosophical analysis to these contested issues by providing an overview of three key topics: What is algorithmic bias? Why and how can it occur? What can and should be done about it? Throughout, we highlight connections—both actual and potential—with philosophical ideas and concerns.

In the hopes of finding supporting evidence for various accounts of actual causation, many philosophers have recently turned to psychological findings about the influence of norms on counterfactual cognition. Surprisingly little philosophical attention has been paid, however, to the question of why considerations of normality should be relevant to counterfactual cognition to begin with. In this paper, I follow two aims. First, against the methodology of two prominent psychological accounts, I argue for a functional approach to understanding the selectivity of counterfactual cognition. Second, I take some steps towards a systematic analysis by providing a qualitative, decision‐theoretic account of one important function of counterfactual thinking, namely, inferring, in the face of undesirable outcomes, corrective policies that prevent the occurrence of similar outcomes in future circumstances. I make a case for employing this analysis by (a) showing its value for assessing the rationality of imagination‐driven counterfactual generation, (b) highlighting its use for making sense of practices in history and policy analysis where counterfactual selection plays a central role, and (c) demonstrating its diagnostic value for identifying areas where counterfactual generation may lead us into epistemic and ethical troubles.

Previous simulation models have found positive effects of cognitive diversity on group performance, but have not explored effects of diversity in demographics (e.g., gender, ethnicity). In this paper, we present an agent-based model that captures two empirically supported hypotheses about how demographic diversity can improve group performance. The results of our simulations suggest that, even when social identities are not associated with distinctive task-related cognitive resources, demographic diversity can, in certain circumstances, benefit collective performance by counteracting two types of conformity that can arise in homogeneous groups: those relating to group-based trust and those connected to normative expectations towards in-groups.

Inspired by recent breakthroughs in predictive modeling, practitioners in both industry and government have turned to machine learning with hopes of operationalizing predictions to drive automated decisions. Unfortunately, many social desiderata concerning consequential decisions, such as justice or fairness, have no natural formulation within a purely predictive framework. In efforts to mitigate these problems, researchers have proposed a variety of metrics for quantifying deviations from various statistical parities that we might expect to observe in a fair world and offered a variety of algorithms in attempts to satisfy subsets of these parities or to trade o the degree to which they are satised against utility. In this paper, we connect this approach to fair machine learning to the literature on ideal and non-ideal methodological approaches in political philosophy. The ideal approach requires positing the principles according to which a just world would operate. In the most straightforward application of ideal theory, one supports a proposed policy by arguing that it closes a discrepancy between the real and the perfectly just world. However, by failing to account for the mechanisms by which our non-ideal world arose, the responsibilities of various decision-makers, and the impacts of proposed policies, naive applications of ideal thinking can lead to misguided interventions. In this paper, we demonstrate a connection between the fair machine learning literature and the ideal approach in political philosophy, and argue that the increasingly apparent shortcomings of proposed fair machine learning algorithms reflect broader troubles faced by the ideal approach. We conclude with a critical discussion of the harms of misguided solutions, a reinterpretation of impossibility results, and directions for future research

It has recently become popular to suggest that cognition can be explained as a process of Bayesian prediction error minimization. Some advocates of this view propose that attention should be understood as the optimization of expected precisions in the prediction-error signal (Clark, 2013, 2016; Feldman & Friston, 2010; Hohwy, 2012, 2013). This proposal successfully accounts for several attention-related phenomena. We claim that it cannot account for all of them, since there are certain forms of voluntary attention that it cannot accommodate. We therefore suggest that, although the theory of Bayesian prediction error minimization introduces some powerful tools for the explanation of mental phenomena, its advocates have been wrong to claim that Bayesian prediction error minimization is ‘all the brain ever does’.