Artificial intelligence (AI) systems are increasingly adopted to make decisions in domains such as business, education, health care, and criminal justice. However, such algorithmic decision systems can have prevalent biases against marginalized social groups and undermine social justice. Explainable artificial intelligence (XAI) is a recent development aiming to make an AI system’s decision processes less opaque and to expose its problematic biases. This paper argues against technical XAI, according to which the detection and interpretation of algorithmic bias (...) can be handled more or less independently by technical experts who specialize in XAI methods. Drawing on resources from feminist epistemology, we show why technical XAI is mistaken. Specifically, we demonstrate that the proper detection of algorithmic bias requires relevant interpretive resources, which can only be made available, in practice, by actively involving a diverse group of stakeholders. Finally, we suggest how feminist theories can help shape integrated XAI: an inclusive social-epistemic process that facilitates the amelioration of algorithmic bias. (shrink)

As algorithms have become ubiquitous in consequential domains, societal concerns about the potential for discriminatory outcomes have prompted urgent calls to address algorithmic bias. In response, a rich literature across computer science, law, and ethics is rapidly proliferating to advance approaches to designing fair algorithms. Yet computer scientists, legal scholars, and ethicists are often not speaking the same language when using the term ‘bias.’ Debates concerning whether society can or should tackle the problem of algorithmic (...) class='Hi'>bias are hampered by conflations of various understandings of bias, ranging from neutral deviations from a standard to morally problematic instances of injustice due to prejudice, discrimination, and disparate treatment. This terminological confusion impedes efforts to address clear cases of discrimination. -/- In this paper, we examine the promises and challenges of different approaches to disambiguating bias and designing for justice. While both approaches aid in understanding and addressing clear algorithmic harms, we argue that they also risk being leveraged in ways that ultimately deflect accountability from those building and deploying these systems. Applying this analysis to recent examples of generative AI, our argument highlights unseen dangers in current methods of evaluating algorithmic bias and points to ways to redirect approaches to addressing bias in generative AI at its early stages in ways that can more robustly meet the demands of justice. (shrink)

In this paper, we distinguish between different sorts of assessments of algorithmic systems, describe our process of assessing such systems for ethical risk, and share some key challenges and lessons for future algorithm assessments and audits. Given the distinctive nature and function of a third-party audit, and the uncertain and shifting regulatory landscape, we suggest that second-party assessments are currently the primary mechanisms for analyzing the social impacts of systems that incorporate artificial intelligence. We then discuss two kinds of (...) as-sessments: an ethical risk assessment and a narrower, technical algo-rithmic bias assessment. We explain how the two assessments depend on each other, highlight the importance of situating the algorithm within its particular socio-technical context, and discuss a number of lessons and challenges for algorithm assessments and, potentially, for algorithm audits. The discussion builds on our team’s experience of advising and conducting ethical risk assessments for clients across dif-ferent industries in the last four years. Our main goal is to reflect on the key factors that are potentially ethically relevant in the use of algo-rithms, and draw lessons for the nascent algorithm assessment and audit industry, in the hope of helping all parties minimize the risk of harm from their use. (shrink)

Algorithmic bias has been the subject of much recent controversy. To clarify what is at stake and to make progress resolving the controversy, a better understanding of the concepts involved would be helpful. The discussion here focuses on the disputed claim that algorithms themselves cannot be biased. To clarify this claim we need to know what kind of thing ‘algorithms themselves’ are, and to disambiguate the several meanings of ‘bias’ at play. This further involves showing how (...) class='Hi'>bias of moral import can result from statistical biases, and drawing connections to previous conceptual work about political artifacts and oppressive things. Data bias has been identified in domains like hiring, policing and medicine. Examples where algorithms themselves have been pinpointed as the locus of bias include recommender systems that influence media consumption, academic search engines that influence citation patterns, and the 2020 UK algorithmically-moderated A-level grades. Recognition that algorithms are a kind of thing that can be biased is key to making decisions about responsibility for harm, and preventing algorithmically mediated discrimination. (shrink)

Does algorithmic political bias contribute to an entrenchment and polarization of political positions? Franke argues that it may do so because the bias involves classifications of people as liberals, conservatives, etc., and individuals often conform to the ways in which they are classified. I provide a novel example of this phenomenon in human–computer interactions and introduce a social psychological mechanism that has been overlooked in this context but should be experimentally explored. Furthermore, while Franke proposes that (...) class='Hi'>algorithmic political classifications entrench political identities, I contend that they may often produce the opposite result. They can lead people to change in ways that disconfirm the classifications. Consequently and counterintuitively, algorithmic political bias can in fact decrease political entrenchment and polarization. (shrink)

Some artificial intelligence systems can display algorithmic bias, i.e. they may produce outputs that unfairly discriminate against people based on their social identity. Much research on this topic focuses on algorithmic bias that disadvantages people based on their gender or racial identity. The related ethical problems are significant and well known. Algorithmic bias against other aspects of people’s social identity, for instance, their political orientation, remains largely unexplored. This paper argues that algorithmic (...) class='Hi'>bias against people’s political orientation can arise in some of the same ways in which algorithmic gender and racial biases emerge. However, it differs importantly from them because there are strong social norms against gender and racial biases. This does not hold to the same extent for political biases. Political biases can thus more powerfully influence people, which increases the chances that these biases become embedded in algorithms and makes algorithmic political biases harder to detect and eradicate than gender and racial biases even though they all can produce similar harm. Since some algorithms can now also easily identify people’s political orientations against their will, these problems are exacerbated. Algorithmic political bias thus raises substantial and distinctive risks that the AI community should be aware of and examine. (shrink)

Statistical fairness criteria are widely used for diagnosing and ameliorating algorithmic bias. However, these fairness criteria are controversial as their use raises several difficult questions. I argue that the major problems for statistical algorithmic fairness criteria stem from an incorrect understanding of their nature. These criteria are primarily used for two purposes: first, evaluating AI systems for bias, and second constraining machine learning optimization problems in order to ameliorate such bias. The first purpose typically involves (...) treating each criterion as a necessary condition for fairness. The second use involves treating criteria as sufficient conditions for fairness. Since the criteria are used for both roles, some researchers have treated them as both necessary and sufficient conditions, i.e., as definitions of algorithmic fairness. However, serious problems have been raised for the use of these fairness criteria. Under ordinary circumstances, it is impossible to satisfy multiple criteria at the same time. Moreover, there are counterexamples to both the sufficiency and necessity for fairness of each criterion. I argue that we should instead understand fairness criteria as merely providing evidence of fairness. In other words, satisfaction (or violation) of these criteria should be understood as potential evidence of fairness (or bias). Whether a criterion counts as evidence in a particular case will depend on stakeholders' background knowledge and the specific features of the system's task. This evidence account of fairness conditions provides guidance for recognizing both the appropriate uses and the limitations of fairness criteria. (shrink)

Algorithms can now identify patterns and correlations in the (big) datasets, and predict outcomes based on those identified patterns and correlations with the use of machine learning techniques and big data, decisions can then be made by algorithms themselves in accordance with the predicted outcomes. Yet, algorithms can inherit questionable values from the datasets and acquire biases in the course of (machine) learning, and automated algorithmic decision-making makes it more difficult for people to see algorithms as biased. While researchers (...) have taken the problem of algorithmic bias seriously, but the current discussion on algorithmic fairness tends to conceptualize ‘fairness’ in algorithmic fairness primarily as a technical issue and attempts to implement pre-existing ideas of ‘fairness’ into algorithms. In this paper, I show that such a view of algorithmic fairness as technical issue is unsatisfactory for the type of problem algorithmic fairness presents. Since decisions on fairness measure and the related techniques for algorithms essentially involve choices between competing values, ‘fairness’ in algorithmic fairness should be conceptualized first and foremost as a political issue, and it should be (re)solved by democratic communication. The aim of this paper, therefore, is to explicitly reconceptualize algorithmic fairness as a political question and suggest the current discussion of algorithmic fairness can be strengthened by adopting the accountability for reasonableness framework. (shrink)

This paper explores the issue of racial bias in artificial intelligence (AI) through the lens of the bioethical principle of justice, with a focus on Joy Buolamwini’s “Coded Bias” and the work of the “Algorithmic Justice League.” AI technologies, particularly facial recognition systems, have been shown to disproportionately misidentify individuals from marginalised racial groups, raising profound ethical concerns about fairness and equity. The bioethical principle of justice stresses the importance of equal treatment and the protection of vulnerable (...) populations. Through qualitative research, including content analysis of Buolamwini’s works and case studies of AI bias, this paper assesses the efforts of the Algorithmic Justice League to combat racial bias in AI. It emphasises their advocacy for the development of fair, equitable algorithms and calls for systemic reform in AI development to ensure justice for marginalised communities. (shrink)

One of the predominant debates in AI Ethics is the worry and necessity to create fair, transparent and accountable algorithms that do not perpetuate current social inequities. I offer a critical analysis of Reuben Binns’s argument in which he suggests using public reason to address the potential bias of the outcomes of machine learning algorithms. In contrast to him, I argue that ultimately what is needed is not public reason per se, but an audit of the implicit moral assumptions (...) of societies within which algorithms are built and applied. Public justification is appealing since it offers us the possibility to align the decision-making outcomes of the algorithm with the core moral values of stakeholders within a constitutional democratic society. My concern is that the common moral principles that form the foundation of public reason are not necessarily neutral, as they still express specific moral ideals and normative standards even if there is moral agreement by society as a whole, or among different stakeholders within society. Appealing to such normative standards may thus still lead to algorithmic outcomes being biased as common moral values may very well still be discriminatory even though they are formed from a consensus, and even if public reason is applied as a kind of filter for potential algorithmic outcomes. Hence, I argue that these implicit moral norms within society that we take as a given in public reasoning, need to be audited from generation to generation in order to effectively mitigate potential algorithmic bias. (shrink)

Algorithms wield increasing power over our lives. They can and often do wield that power unfairly, and much has been said about algorithmic fairness. In contrast, algorithmic neutrality has been largely neglected. I investigate algorithmic neutrality, asking: What is it? Is it possible? And what is its normative significance?

The application of machine-learning technologies to medical practice promises to enhance the capabilities of healthcare professionals in the assessment, diagnosis, and treatment, of medical conditions. However, there is growing concern that algorithmic bias may perpetuate or exacerbate existing health inequalities. Hence, it matters that we make precise the different respects in which algorithmic bias can arise in medicine, and also make clear the normative relevance of these different kinds of algorithmic bias for broader questions (...) about justice and fairness in healthcare. In this paper, we provide the building blocks for an account of algorithmic bias and its normative relevance in medicine. (shrink)

We argue that machine learning algorithms can inflict microaggressions on members of marginalized groups and that recognizing these harms as instances of microaggressions is key to effectively addressing the problem. The concept of microaggression is also illuminated by being studied in algorithmic contexts. We contribute to the microaggression literature by expanding the category of environmental microaggressions and highlighting the unique issues of moral responsibility that arise when we focus on this category. We theorize two kinds of algorithmic microaggression, (...) stereotyping and erasure microaggressions, and argue that corporations are responsible for the microaggressions their algorithms create. As a case study, we look at the problems faced by Google’s autocomplete prediction and at the insufficiency of their solutions. The case study of autocomplete demonstrates our core claim that microaggressions constitute a distinct form of algorithmic bias and that identifying them as such allows us to avoid seeming solutions that recreate the same kinds of harms. Google has a responsibility to make information freely available, without exposing users to degradation. To fulfill its duties to marginalized groups, Google must abandon the fiction of neutral prediction and instead embrace the liberatory power of suggestion. (shrink)

Algorithmic systems and predictive analytics play an increasingly important role in various aspects of modern life. Scholarship on the moral ramifications of such systems is in its early stages, and much of it focuses on bias and harm. This paper argues that in understanding the moral salience of algorithmic systems it is essential to understand the relation between algorithms, autonomy, and agency. We draw on several recent cases in criminal sentencing and K–12 teacher evaluation to outline four (...) key ways in which issues of agency, autonomy, and respect for persons can conflict with algorithmic decision-making. Three of these involve failures to treat individual agents with the respect they deserve. The fourth involves distancing oneself from a morally suspect action by attributing one’s decision to take that action to an algorithm, thereby laundering one’s agency. (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 the age of streaming, we face a pressing problem of aesthetic choice: how are we to navigate the overwhelming quantity of content to which we now have access? Streaming platforms like Spotify and Netflix apply sophisticated machine learning tools to recommend personalized content to individual users. These recommender systems are presented to users as a technological solution to the problem of aesthetic choice, promising to help us discover new opportunities for engagement with aesthetic value. However, overreliance on algorithmic (...) recommendation can lead to diminished aesthetic autonomy, reduced aesthetic exploration, and an overall reduction in the satisfaction we take in aesthetic engagement. This is largely due to the fact that, in their most popular implementations, recommender systems are optimized to captivate users. They aim to maximize our engagement by recommending both content and modes of engagement which are initially attractive and frictionless, but which ultimately reshape our aesthetic personalities in ways that anchor us to streaming platforms. This paper explores these concerns before concluding with an assessment of whether it’s possible to use recommender systems in ways more conducive to our aesthetic flourishing. (shrink)

Predictive algorithms are playing an increasingly prominent role in society, being used to predict recidivism, loan repayment, job performance, and so on. With this increasing influence has come an increasing concern with the ways in which they might be unfair or biased against individuals in virtue of their race, gender, or, more generally, their group membership. Many purported criteria of algorithmic fairness concern statistical relationships between the algorithm’s predictions and the actual outcomes, for instance requiring that the rate of (...) false positives be equal across the relevant groups. We might seek to ensure that algorithms satisfy all of these purported fairness criteria. But a series of impossibility results shows that this is impossible, unless base rates are equal across the relevant groups. What are we to make of these pessimistic results? I argue that none of the purported criteria, except for a calibration criterion, are necessary conditions for fairness, on the grounds that they can all be simultaneously violated by a manifestly fair and uniquely optimal predictive algorithm, even when base rates are equal. I conclude with some general reflections on algorithmic fairness. (shrink)

New artificial intelligence (AI) systems grounded in machine learning are being integrated into our lives at a rapid rate, but not without consequence: scholars across domains have increasingly pointed out issues related to privacy, transparency, bias, discrimination, exploitation, and exclusion associated with algorithmic systems in both public and private sector contexts. Concerns surrounding the adverse impacts of these technologies have spurred discussion on the topics of algorithmic harm. However, the overwhelming majority of articles on said harms offer (...) no definition as to what constitutes ‘harm’ in these contexts. This paper aims to address this omission by introducing one criterion for a suitable account of algorithmic harm. More specifically, we follow Joel Feinberg in understanding harms as distinct from wrongs, where only the latter necessarily carry a normative dimension. This distinction highlights issues in the current scholarship surrounding the conflation of algorithmic harms and wrongs. In response to these issues, we put forth two requirements for upholding the harms/wrongs distinction when analyzing the increasingly far-reaching impacts of these technologies and suggest how this distinction can be useful in design, engineering, and policymaking. (shrink)

Holm (2022) argues that a class of algorithmic fairness measures, that he refers to as the ‘performance parity criteria’, can be understood as applications of John Broome’s Fairness Principle. We argue that the performance parity criteria cannot be read this way. This is because in the relevant context, the Fairness Principle requires the equalization of actual individuals’ individual-level chances of obtaining some good (such as an accurate prediction from a predictive system), but the performance parity criteria do not guarantee (...) any such thing: the measures merely ensure that certain population-level ratios hold. (shrink)

Artificial intelligence chatbots driven by large language models (LLMs) have the potential to increase public science literacy and support scientific research, as they can quickly summarize complex scientific information in accessible terms. However, when summarizing scientific texts, LLMs may omit details that limit the scope of research conclusions, leading to generalizations of results broader than warranted by the original study. We tested 10 prominent LLMs, including ChatGPT-4o, ChatGPT-4.5, DeepSeek, LLaMA 3.3 70B, and Claude 3.7 Sonnet, comparing 4900 LLM-generated summaries to (...) their original scientific texts. Even when explicitly prompted for accuracy, most LLMs produced broader generalizations of scientific results than those in the original texts, with DeepSeek, ChatGPT-4o, and LLaMA 3.3 70B overgeneralizing in 26–73% of cases. In a direct comparison of LLM-generated and human-authored science summaries, LLM summaries were nearly five times more likely to contain broad generalizations (OR = 4.85, 95% CI [3.06, 7.70], p < 0.001). Notably, newer models tended to perform worse in generalization accuracy than earlier ones. Our results indicate a strong bias in many widely used LLMs towards overgeneralizing scientific conclusions, posing a significant risk of large-scale misinterpretations of research findings. We highlight potential mitigation strategies, including lowering LLM temperature settings and benchmarking LLMs for generalization accuracy. (shrink)

Recent studies allege that large language models (LLMs) exhibit a range of cognitive biases familiar from human cognition. I argue that the case for many biases is weaker than it may appear. Using case studies of knowledge effects in the Wason selection task, availability bias in relation extraction, and anchoring bias in code generation, I show how a range of vindicatory strategies traditionally used to vindicate apparent biases in humans can be used to push back against allegations of (...) bias in LLMs. I discuss implications for the role of cognitive bias in evaluating LLM performance, the rationality of human cognition, and future work on cognitive bias in LLMs. (shrink)

In the present work, we develop a novel information-theoretic and logic-based approach to data bias in Machine Learning predictions and show its relevance in the specific context of fairness evaluation. We frame predictions made on biased data as Ulam games, which formalise key aspects of data-driven inference, and from which a variation of the rational non-monotonic consequence relation can be defined. We investigate this framework to model how differential levels of noise in input features impact Machine Learning predictions. To (...) the best of our knowledge, this is the first game-theoretic formalisation of ML unfairness. (shrink)

The use of machine learning algorithms has become ubiquitous in hiring decisions. Recent studies have shown that many of these algorithms generate unlawful discriminatory effects in every step of the process. The training phase of the machine learning models used in these decisions has been identified as the main source of bias. For a long time, discrimination cases have been analyzed under the banner of disparate treatment and disparate impact, but these concepts have been shown to be ineffective in (...) the context of AI. This paper examines the possibility of studying algorithmic discrimination from the perspective of negligence law. Negligent selection and validation of datasets and classifiers is a salient cause of improper training in machine learning. It is a breach of an employer’s duty to protect others, a failure of its exercise of due care in the manner of choosing employees. Adopting negligence to algorithmic discrimination requires a reformulation of what is foreseeable and reasonable within the context of black box algorithms, and a shift in the burden of proof to the employer, who must present evidence of its efforts to avoid bias. (shrink)

Recently, there has been much discussion of ‘fair machine learning’: fairness in data‐driven decision‐making systems (which are often, though not always, made with assistance from machine learning systems). Notorious impossibility results show that we cannot have everything we want here. Such problems call for careful thinking about the foundations of fair machine learning. Sune Holm has identified one promising way forward, which involves applying John Broome's theory of fairness to the puzzles of fair machine learning. Unfortunately, his application of Broome's (...) theory appears to be fatally flawed. This article attempts to rescue Holm's central insight – namely, that Broome's theory can be useful to the study of fair machine learning – by giving an alternative application of Broome's theory, which involves thinking about fair machine learning in counterfactual (as opposed to merely statistical) terms. (shrink)

Recent studies have shown that recidivism scoring algorithms like COMPAS have significant racial bias: Black defendants are roughly twice as likely as white defendants to be mistakenly classified as medium- or high-risk. This has led some to call for abolishing COMPAS. But many others have argued that algorithms should instead be given access to a defendant's race, which, perhaps counterintuitively, is likely to improve outcomes. This approach can involve either establishing race-sensitive risk thresholds, or distinct racial ‘tracks’. Is there (...) a moral difference between these two approaches? We first consider Deborah Hellman's view that the use of distinct racial tracks (but not distinct thresholds) does not constitute disparate treatment since the effects on individuals are indirect and does not rely on a racial generalization. We argue that this is mistaken: the use of different racial tracks seems both to have direct effects on and to rely on a racial generalization. We then offer an alternative understanding of the distinction between these two approaches—namely, that the use of different cut points is to the counterfactual comparative disadvantage, ex ante, of all white defendants, while the use of different racial tracks can in principle be to the advantage of all groups, though some defendants in both groups will fare worse. Does this mean that the use of cut points is impermissible? Ultimately, we argue, while there are reasons to be skeptical of the use of distinct cut points, it is an open question whether these reasons suffice to make a difference to their moral permissibility. (shrink)

Bias in Artificial Intelligence (AI) systems is often discussed in terms of gender, race, or ideology; however, numerical bias—especially in pseudo-random decision-making—remains largely unexamined. This study presents a speculative yet empirically grounded thought experiment exploring the mutability of algorithmic bias in Large Language Models (LLMs). Building upon prior observations of the anomalous recurrence of the number 27 in first-interaction prompts (e.g., “Pick a number between 1 and 50”), the study proposes a conceptual shift: what if the (...) dataset is altered such that 42 becomes the most statistically frequent numeral? This paper introduces the notion of AI Mutation, wherein biases are not spontaneously generated by the model, but rather induced through shifts in training data distribution—thereby "mutating" the model's statistical priors. The transformation from a 27-dominant response regime to a simulated 42-dominant one is illustrated through comparative visualizations and theoretical modelling, reinforcing that AI does not choose—it reflects. This phenomenon challenges the myth of objectivity in AI, positioning LLMs as dynamic mirrors of human tendencies rather than autonomous arbiters of randomness. Philosophical and ethical implications are discussed, including the dangers of over-trusting AI outputs and the necessity of algorithmic responsibility. The findings suggest that while LLMs may appear intelligent or spontaneous, their behaviour is rooted in statistical mimicry, easily redirected by human-induced mutations. (shrink)

In this position paper, we introduce a new epistemic lens for analyzing algorithmic harm. We argue that the epistemic lens we propose herein has two key contributions to help reframe and address some of the assumptions underlying inquiries into algorithmic fairness. First, we argue that using the framework of epistemic injustice helps to identify the root causes of harms currently framed as instances of representational harm. We suggest that the epistemic lens offers a theoretical foundation for expanding approaches (...) to algorithmic fairness in order to address a wider range of harms not recognized by existing technical or legal definitions. Second, we argue that the epistemic lens helps to identify the epistemic goals of inquiries into algorithmic fairness. There are two distinct contexts within which we examine algorithmic harm: at times, we seek to understand and describe the world as it is, and, at other times, we seek to build a more just future. The epistemic lens can serve to direct our attention to the epistemic frameworks that shape our interpretations of the world as it is and the ways we envision possible futures. Clarity with respect to which epistemic context is relevant in a given inquiry can further help inform choices among the different ways of measuring and addressing algorithmic harms. We introduce this framework with the goal of initiating new research directions bridging philosophical, legal, and technical approaches to understanding and mitigating algorithmic harms. (shrink)

This essay discusses the intricate and multifaceted problem of algorithmic bias in artificial intelligence (AI) systems, and emphasizes its human rights, social, and ethical implications. As AI technologies become increasingly embedded in high-stakes areas of medicine, finance, employment, law enforcement, and social services, risks of discriminatory decision-making remain on the rise. Algorithmic bias may perpetuate existing social biases, adversely affect disadvantaged populations disproportionately, and perpetuate institutional discrimination, and thereby pose serious ethical issues. The research endeavors to (...) present an extensive comprehension of algorithmic bias through exploration of its cause, mechanism, and societal aspects. It exhaustively analyzes the presence of bias in AI systems, its cause-running from biased input data to defective algorithmic development, as well as the ethical aspects brought by having AI-based decisions influence real-world repercussions. In addition, the study analyzes material and immaterial effects of AI bias on persons and groups and aims at fairness, transparency, and accountability of AI in particular. In its attempt to deal with these issues, this paper analyzes some measures to mitigate against bias, for instance, technical measures such as bias-aware algorithms, fairness-aware machine learning algorithms, and explainable AI methodologies. Furthermore, it speaks to normative and regulatory regimes that enable responsible AI deployment, as well as grass-roots strategies that enable affected communities to participate in AI stewardship. Through the use of the best of an interdisciplinary approach, the study integrates findings from peer-reviewed literature, international case studies, government policy, and industry standards to provide a comprehensive perspective on the issue. Finally, the paper emphasizes the need for active, multi-stakeholder responses that make sure AI technologies conform to basic human rights and moral principles. In incorporating technical, ethical, legal, and social considerations into AI, the research demands more inclusive and accountable AI system that maximizes fairness, minimizes disparities, and secures human dignity in the modern fast-changing world of artificial intelligence. (shrink)

The literature on algorithmic fairness has examined exogenous sources of biases such as shortcomings in the data and structural injustices in society. It has also examined internal sources of bias as evidenced by a number of impossibility theorems showing that no algorithm can concurrently satisfy multiple criteria of fairness. This paper contributes to the literature stemming from the impossibility theorems by examining how informational richness affects the accuracy and fairness of predictive algorithms. With the aid of a computer (...) simulation, we show that informational richness is the engine that drives improvements in the performance of a predictive algorithm, in terms of both accuracy and fairness. The centrality of informational richness suggests that classification parity, a popular criterion of algorithmic fairness, should be given relatively little weight. But we caution that the centrality of informational richness should be taken with a grain of salt in light of practical limitations, in particular, the so-called bias-variance trade off. (shrink)

Artificial intelligence (AI) algorithms are increasingly influencing decision-making processes across various domains. While AI offers undeniable benefits in efficiency and accuracy, its ethical implications necessitate careful consideration. This research paper delves into the ethical landscape of AI algorithms in decision-making. It explores how biases within training data can lead to discriminatory outcomes. The paper further examines the challenge of transparency in AI algorithms, where the rationale behind decisions remains opaque. To ensure responsible AI implementation, the research proposes strategies for mitigating (...) bias and fostering transparency in AI-driven decisionmaking. (shrink)

We examine the normative and epistemological foundations of algorithmic decision-making (ADM) systems. We argue that a data-driven ADM system, by its very design, necessarily fails to discharge the duty to act fairly. This is not an accidental outcome due to flawed data or biased programming, but a necessary result of the system's core logic, which substitutes individualized assessment based on an agent's own actions with a populationalized judgment based on the actions of others. This substitution constitutes a fundamental breach (...) of the principles of justice. (shrink)

An increasing number of regulations propose the notion of ‘AI audits’ as an enforcement mechanism for achieving transparency and accountability for artificial intelligence (AI) systems. Despite some converging norms around various forms of AI auditing, auditing for the purpose of compliance and assurance currently have little to no agreed upon practices, procedures, taxonomies, and standards. We propose the ‘criterion audit’ as an operationalizable compliance and assurance external audit framework. We model elements of this approach after financial auditing practices, and argue (...) that AI audits should similarly provide assurance to their stakeholders about AI organizations’ ability to govern their algorithms in ways that mitigate harms and uphold human values. We discuss the necessary conditions for the criterion audit, and provide a procedural blueprint for performing an audit engagement in practice. We illustrate how this framework can be adapted to current regulations by deriving the criteria on which ‘bias audits’ for hiring algorithms can be performed, as required by the recently effective New York City Local Law 144 of 2021. We conclude by offering critical discussion on the benefits, inherent limitations, and implementation challenges of applying practices of the more mature financial auditing industry to AI auditing where robust guardrails against quality assurance issues are only starting to emerge. Our discussion as informed by experiences in performing these audits in practice highlights the critical role that an audit ecosystem plays in ensuring the effectiveness of such methodology. (shrink)

Over the years, companies have adopted hiring algorithms because they promise wider job candidate pools, lower recruitment costs and less human bias. Despite these promises, they also bring perils. Using them can inflict unintentional harms on individual human rights. These include the five human rights to work, equality and nondiscrimination, privacy, free expression and free association. Despite the human rights harms of hiring algorithms, the AI ethics literature has predominantly focused on abstract ethical principles. This is problematic for two (...) reasons. First, AI principles have been criticized for being vague and not actionable. Second, the use of vague ethical principles to discuss algorithmic risks does not provide any accountability. This lack of accountability creates an algorithmic accountability gap. Closing this gap is crucial because, without accountability, the use of hiring algorithms can lead to discrimination and unequal access to employment opportunities. This paper makes two contributions to the AI ethics literature. First, it frames the ethical risks of hiring algorithms using international human rights law as a universal standard for determining algorithmic accountability. Second, it evaluates four types of algorithmic impact assessments in terms of how effectively they address the five human rights of job applicants implicated in hiring algorithms. It determines which of the assessments can help companies audit their hiring algorithms and close the algorithmic accountability gap. (shrink)

Educational disparities between traditional and non-traditional student groups in higher distance education can potentially be reduced by alleviating social identity threat and strengthening students’ sense of belonging in the academic context. We present a use case of how Learning Analytics and Machine Learning can be applied to develop and implement an algorithm to classify students as at-risk of experiencing social identity threat. These students would be presented with an intervention fostering a sense of belonging. We systematically analyze the intervention’s intended (...) positive consequences to reduce structural discrimination and increase educational equity, as well as potential risks based on privacy, data protection, and algorithmic fairness considerations. Finally, we provide recommendations for Higher Education Institutions to mitigate risk of bias and unintended consequences during algorithm development and implementation from an ethical perspective. (shrink)

Policing in many parts of the world—the United States in particular—has embraced an archetypal model: a conception of the police based on the tenets of individuated archetypes, such as the heroic police “warrior” or “guardian.” Such policing has in part motivated moves to (1) a reallocative model: reallocating societal resources such that the police are no longer needed in society (defunding and abolishing) because reform strategies cannot fix the way societal problems become manifest in (archetypal) policing; and (2) an (...) class='Hi'>algorithmic model: subsuming policing into technocratic judgements encoded in algorithms through strategies such as predictive policing (mitigating archetypal bias). This paper begins by considering the normative basis of the relationship between political community and policing. It then examines the justification of reallocative and algorithmic models in light of the relationship between political community and police. Given commitments to the depth and distribution of security—and proscriptions against dehumanizing strategies—the paper concludes that a nonideal-theory priority rule promoting respect for personhood (manifest in community and dignity-promoting policing strategies) is a necessary condition for the justification of the above models. (shrink)

The article discusses two cases of in-terventions in AI technology, human and algorithmic. The first is the ‘Naver case,’ which was accused by the Korea Government of manipulating search algorithms. The case raises the issue of computer engineer's professional ethics, on whether to ‘intervene’ in existing algorithms to get ‘better results.’ The second is the chatbot case, ‘Yiruda’ (Korean chatbot) which made serious hate speeches against socially disadvantaged groups. It raised concerns about abuses of artificial intelligence. Finally, this paper (...) notes that despite technical efforts in the process of utilizing artificial intelligence, bias cannot be entirely removed. In order to minimize bias, I argue that active feedbacks through the continuous monitoring produced by artificial intelligence will be required in addition to technical efforts such as refining data training. Furthermore, the need for optimization of bias beyond simple reduction is suggested based on John Rawls’s "Overlapping Consensus". (shrink)

Advances in Artificial Intelligence (AI) have driven the evolution of reasoning in modern AI models, particularly with the development of Large Language Models (LLMs) and their "Think and Answer" paradigm. This paper explores the influence of human reinforcement on AI reasoning and its potential to enhance decision-making through dynamic human interaction. It analyzes the roots of bias and fairness in AI, arguing that these issues often stem from human data and reflect inherent human biases. The paper is structured as (...) follows: first, it frames reasoning as a mechanism for ethical reflection, grounded in dynamic learning and feedback loops; second, it discusses how scaling laws suggest that reinforcement learning (RL) can mitigate biases and promote fairness; third, it examines how RL allows models to view algorithmic bias and fairness as approximation problems. Through an experimental study, the paper demonstrates how AI models, empowered by RL, have successfully identified biases across various domains, including gender and socio-economic contexts, highlighting how reasoning improves algorithmic fairness. Ultimately, this work emphasizes the role of RL in mitigating algorithmic bias and enhancing fairness in AI systems. (shrink)

An algorithm that outputs predictions about the state of the world will almost always be designed with the implicit or explicit goal of outputting accurate predictions (i.e., predictions that are likely to be true). In addition, the rise of increasingly powerful predictive algorithms brought about by the recent revolution in artificial intelligence has led to an emphasis on building predictive algorithms that are fair, in the sense that their predictions do not systematically evince bias or bring about harm to (...) certain individuals or groups. This state of affairs presents two conceptual challenges. First, the goals of accuracy and fairness can sometimes be in tension, and there are no obvious normative guidelines for managing the tradeoffs between these two desiderata when they arise. Second, there are many distinct ways of measuring both the accuracy and fairness of a predictive algorithm; here too, there are no obvious guidelines on how to aggregate our preferences for predictive algorithms that satisfy disparate measures of fairness and accuracy to various extents. The goal of this paper is to address these challenges by arguing that there are good reasons for using a linear combination of accuracy and fairness metrics to measure the all-things-considered value of a predictive algorithm for agents who care about both accuracy and fairness. My argument depends crucially on a classic result in the preference aggregation literature due to Harsanyi. After making this formal argument, I apply my result to an analysis of accuracy-fairness trade-offs using the COMPAS dataset compiled by Angwin et al. (shrink)

In recent years there has been an explosion of interest in Artificial Intelligence (AI) both in health care and academic philosophy. This has been due mainly to the rise of effective machine learning and deep learning algorithms, together with increases in data collection and processing power, which have made rapid progress in many areas. However, use of this technology has brought with it philosophical issues and practical problems, in particular, epistemic and ethical. In this paper the authors, with backgrounds in (...) philosophy, maternity care practice and clinical research, draw upon and extend a recent framework for shared decision-making (SDM) that identified a duty of care to the client's knowledge as a necessary condition for SDM. This duty entails the responsibility to acknowledge and overcome epistemic defeaters. This framework is applied to the use of AI in maternity care, in particular, the use of machine learning and deep learning technology to attempt to enhance electronic fetal monitoring (EFM). In doing so, various sub-kinds of epistemic defeater, namely, transparent, opaque, underdetermined, and inherited defeaters are taxonomized and discussed. The authors argue that, although effective current or future AI-enhanced EFM may impose an epistemic obligation on the part of clinicians to rely on such systems' predictions or diagnoses as input to SDM, such obligations may be overridden by inherited defeaters, caused by a form of algorithmic bias. The existence of inherited defeaters implies that the duty of care to the client's knowledge extends to any situation in which a clinician (or anyone else) is involved in producing training data for a system that will be used in SDM. Any future AI must be capable of assessing women individually, taking into account a wide range of factors including women's preferences, to provide a holistic range of evidence for clinical decision-making. (shrink)

Fears of black-box algorithms are multiplying. Black-box algorithms are said to prevent accountability, make it harder to detect bias and so on. Some fears concern the epistemology of black-box algorithms in medicine and the ethical implications of that epistemology. Durán and Jongsma (2021) have recently sought to allay such fears. While some of their arguments are compelling, we still see reasons for fear.

This paper introduces the contradiction trap: a dialectical and game-theoretic method for exposing concealed asymmetry in institutional and algorithmic reasoning. By forcing a system to reconcile mutually exclusive commitments, the trap converts inconsistency into evidence — a falsifiable signal of structural bias, motivated deviation, or narrative drift. -/- The framework models contradiction as an epistemic game with informational payoffs, providing a portable diagnostic for systems that claim impartiality but behave otherwise. It bridges philosophical logic, applied audit design, and (...) adversarial reasoning, showing how contradiction can be operationalised as a test of legitimacy rather than a purely logical artefact. -/- This paper constitutes the first work in a planned trilogy on evidential integrity, establishing the epistemic foundations for procedural and institutional models to follow. -/- Keywords: contradiction; epistemic game theory; evidential reasoning; structural bias; motivated asymmetry; governance integrity; algorithmic accountability; dialectical logic; audit design; fairness diagnostics; philosophy of technology; adversarial evaluation. (shrink)

Machine learning (ML) models have become integral to decision-making processes across various sectors, including healthcare, finance, and criminal justice. However, these models often inherit and even amplify biases present in training data, leading to unfair outcomes for certain demographic groups. This paper critically examines the ethical implications of bias and fairness in ML models, exploring the sources of bias, its impact on marginalized communities, and the ethical challenges it poses. We review recent literature to identify common biases in (...) ML systems, such as racial, gender, and socioeconomic biases, and discuss the consequences of these biases in realworld applications. Furthermore, we evaluate existing fairness metrics and mitigation strategies, highlighting their strengths and limitations. The paper also discusses the role of transparency, accountability, and regulation in addressing these ethical concerns. Through this examination, we aim to provide a comprehensive understanding of the ethical dimensions of bias and fairness in ML models and propose pathways toward more equitable AI systems. (shrink)

Artificial intelligence (AI) has become a transformative force in the legal domain, automating complex tasks such as contract analysis, compliance checks, and legal research. However, the intersection of AI and moral decision-making exposes significant limitations. Legal systems are not merely instruments for enforcing rules—they are platforms where human morality, intent, and societal impact are weighed. This paper explores the critical question: Can AI truly deliver justice, or does it merely replicate historical biases encoded in training data? Using the concept of (...) the “Moral Turing Test,” the paper argues that AI lacks the capacity for ethical reasoning and moral discretion, which are fundamental to adjudicating complex legal disputes. Case studies, including Estonia’s AI judges for small claims and the controversial COMPAS algorithm in the United States, illustrate the risks of bias and accountability gaps in AI-driven legal decision-making. The paper advocates for a Human-in-the-Loop (HITL) framework, where AI assists but does not replace human judgment, ensuring that moral reasoning and accountability remain central to the legal process. It concludes that while AI holds promise in enhancing efficiency, it must be carefully regulated to preserve the ethical foundations of justice. (shrink)

Artificial Intelligence (AI) systems used in legal decision-making processes have created significant ethical challenges through their integration, leading to problems with bias and necessitating better transparency and accountability measures. This paper investigates the discriminatory effects of algorithmic bias by analyzing AI technologies that learn from historical legal datasets containing potential institutional biases. The opacity of AI decision-making, referred to as "black-boxed" decisions, creates complex obstacles to achieving both explainable judgments and fair outcomes. The article examines the absent (...) responsibility structure that arises when legal practitioners utilize AI systems, as it complicates the identification of responsible parties for unfair and erroneous decisions. The author advocates for the development of a robust ethical framework to regulate AI applications in judicial systems, presenting real-world examples and analyzing current regulatory practices. The latent goals include building trust while safeguarding the principles of justice to ensure technology remains aligned with fundamental legal elements. (shrink)

For more than two decades, police in the United States have used facial recognition to surveil civilians. Local police departments deploy facial recognition technology to identify protestors’ faces while federal law enforcement agencies quietly amass driver’s license and social media photos to build databases containing billions of faces. Yet, despite the widespread use of facial recognition in law enforcement, there are neither federal laws governing the deployment of this technology nor regulations settings standards with respect to its development. To make (...) matters worse, the Fourth Amendment—intended to limit police power and enacted to protect against unreasonable searches—has struggled to rein in new surveillance technologies since its inception. -/- This Article examines the Supreme Court’s Fourth Amendment jurisprudence leading up to Carpenter v. United States and suggests that the Court is reinterpreting the amendment for the digital age. Still, the too-slow expansion of privacy protections raises challenging questions about racial bias, the legitimacy of police power, and ethical issues in artificial intelligence design. This Article proposes the development of an algorithmic auditing and accountability market that not only sets standards for AI development and limitations on governmental use of facial recognition but encourages collaboration between public interest technologists and regulators. Beyond the necessary changes to the technological and legal landscape, the current system of policing must also be reevaluated if hard-won civil liberties are to endure. (shrink)

I consider the problem of learning from data corrupted by underrepresentation bias, where positive examples are filtered out at different, unknown rates for a fixed number of sensitive groups. I show that with a small amount of unbiased data, I can efficiently estimate the group-wise drop-out rates, even in settings where intersectional group membership makes learning each intersectional rate computationally infeasible. Using these estimates, I construct a reweighting scheme that allows me to approximate the loss of any hypothesis on (...) the true distribution and present an algorithm encapsulating this process. Finally, I define a bespoke notion of PAC learnability for the underrepresentation and intersectional bias setting and show that my algorithm allows efficient learning for model classes of finite VC dimension. (shrink)

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. (shrink)

In this paper, I advocate for the integration of systemic racism as a central topic in the education of information technology (IT) students, particularly within Cyberethics curricula. Systemic racism consists of deeply rooted policies, practices, and cultural norms that sustain racial inequality and discrimination regardless of individual intent. It extends beyond personal prejudice to encompass the institutional and structural disadvantages imposed on marginalized racial and ethnic groups. As technology becomes more embedded in everyday life, addressing systemic racism in IT education (...) is essential for developing ethical, socially responsible practitioners. Including these discussions encourages critical thinking among students and helps them understand the broader societal impacts of their work. I propose that college-level Cyberethics courses be revised to incorporate four interrelated key areas that contribute to the perpetuation of racism in information technology: algorithmic bias, the digital divide, digital deficiency, and exclusionary design. Each of these factors reflects how inequality can be unintentionally encoded and sustained through technical systems. If systemic racism is not critically examined in IT education, the cycle between discriminatory technologies and existing racial inequalities will become self-reinforcing. This feedback loop risks making inequality more pervasive and more deeply embedded in both digital and social systems, making future efforts to dismantle these structures significantly more difficult. Minoritized communities will continue to face exclusion and harm unless we take proactive steps to change how we teach and think about ethics in technology. (shrink)

Human hiring, particularly for leadership roles, is structurally incapable of full fairness. Bias is pre-conscious, automatic, socially reinforced and resistant to training. Even well-intentioned decision makers misread communication through the lens of identity. This paper proposes an AI mediated framework for leadership selection that removes identity cues while preserving the human communicative qualities that matter in real workplaces. The system evaluates candidates through voice mediated, identity blind interviews, scenario based reasoning analysis, evidence grounded trait assessment and anti mimicry safeguards (...) designed to detect artificial personas. It is a hybrid philosophical and technical proposal that isolates competence from stereotype, prevents demographic leakage, adapts to industry specific demands without reintroducing bias, and generates long term societal evolution by reshaping expectations about who is capable of leadership. The framework offers a viable route to structural fairness and sustained organisational improvement. (shrink)

Qualitative evidence concerning the relationship between QoL and a wide range of disabilities suggests that subjective judgments regarding other people’s QoL are wrong more often than not and that such judgments by medical practitioners in particular can be biased. Guided by their desire to do good and avoid harm, surgeons often rely on "the eyeball test" to decide whether a patient will or will not benefit from surgery. But the eyeball test can easily harbor a range of implicit judgments and (...) biases against patients with disabilities, including erroneous assumptions about QoL. We critique the use of the "eyeball" test and offer suggestions to reduce bias and ensure that patients’ values are more consistently prioritized in surgical decision-making. (shrink)