Research Papers

What are People Talking about in #BlackLivesMatter and #StopAsianHate? Exploring and Categorizing Twitter Topics Emerging in Online Social Movements through the Latent Dirichlet Allocation Model

Xin Tong, Yixuan Li, Jiayi Li +2

Minority groups have been using social media to organize social movements that create profound social impacts. Black Lives Matter (BLM) and Stop Asian Hate (SAH) are two successful social movements that have spread on Twitter that promote protests and activities against racism and increase the public's awareness of other social challenges that minority groups face. However, previous studies have mostly conducted qualitative analyses of tweets or interviews with users, which may not comprehensively and validly represent all tweets. Very few studies have explored the Twitter topics within BLM and SAH dialogs in a rigorous, quantified and data-centered approach. Therefore, in this research, we adopted a mixed-methods approach to comprehensively analyze BLM and SAH Twitter topics. We implemented (1) the latent Dirichlet allocation model to understand the top high-level words and topics and (2) open-coding analysis to identify specific themes across the tweets. We collected more than one million tweets with the #blacklivesmatter and #stopasianhate hashtags and compared their topics. Our findings revealed that the tweets discussed a variety of influential topics in depth, and social justice, social movements, and emotional sentiments were common topics in both movements, though with unique subtopics for each movement. Our study contributes to the topic analysis of social movements on social media platforms in particular and the literature on the interplay of AI, ethics, and society in general.

Many researchers and policymakers have expressed excitement about algorithmic explanations enabling more fair and responsible decision-making. However, recent experimental studies have found that explanations do not always improve human use of algorithmic advice. In this study, we shed light on how people interpret and respond to counterfactual explanations (CFEs) -- explanations that show how a model's output would change with marginal changes to its input(s) -- in the context of pretrial risk assessment instruments (PRAIs). We ran think-aloud trials with eight sitting U.S. state court judges, providing them with recommendations from a PRAI that includes CFEs. We found that the CFEs did not alter the judges' decisions. At first, judges misinterpreted the counterfactuals as real -- rather than hypothetical -- changes to defendants. Once judges understood what the counterfactuals meant, they ignored them, stating their role is only to make decisions regarding the actual defendant in question. The judges also expressed a mix of reasons for ignoring or following the advice of the PRAI without CFEs. These results add to the literature detailing the unexpected ways in which people respond to algorithms and explanations. They also highlight new challenges associated with improving human-algorithm collaborations through explanations.

Algorithmic Bias and Data Bias: Understanding the Relation between Distributionally Robust Optimization and Data Curation

Agnieszka Słowik, Léon Bottou

Machine learning systems based on minimizing average error have been shown to perform inconsistently across notable subsets of the data, which is not exposed by a low average error for the entire dataset. In consequential social and economic applications, where data represent people, this can lead to discrimination of underrepresented gender and ethnic groups. Given the importance of bias mitigation in machine learning, the topic leads to contentious debates on how to ensure fairness in practice (data bias versus algorithmic bias). Distributionally Robust Optimization (DRO) seemingly addresses this problem by minimizing the worst expected risk across subpopulations. We establish theoretical results that clarify the relation between DRO and the optimization of the same loss averaged on an adequately weighted training dataset. The results cover finite and infinite number of training distributions, as well as convex and non-convex loss functions. We show that neither DRO nor curating the training set should be construed as a complete solution for bias mitigation: in the same way that there is no universally robust training set, there is no universal way to setup a DRO problem and ensure a socially acceptable set of results. We then leverage these insights to provide a mininal set of practical recommendations for addressing bias with DRO. Finally, we discuss ramifications of our results in other related applications of DRO, using an example of adversarial robustness. Our results show that there is merit to both the algorithm-focused and the data-focused side of the bias debate, as long as arguments in favor of these positions are precisely qualified and backed by relevant mathematics known today.

While deep learning (DL) approaches are reaching human-level performance for many tasks, including for diagnostics AI, the focus is now on challenges possibly affecting DL deployment, including AI privacy, domain generalization, and fairness. This last challenge is addressed in this study. Here we look at a novel method for ensuring AI fairness with respect to protected or sensitive factors. This method uses domain adaptation via training set enhancement to tackle bias-causing training data imbalance. More specifically, it uses generative models that allow the generation of more synthetic training samples for underrepresented populations. This paper applies this method to the use case of detection of age related macular degeneration (AMD). Our experiments show that starting with an originally biased AMD diagnostics model the method has the ability to improve fairness.

We propose a novel method for enforcing AI fairness with respect to protected or sensitive factors. This method uses a dual strategy performing training and representation alteration (TARA) for the mitigation of prominent causes of AI bias by including: a) the use of representation learning alteration via adversarial independence to suppress the bias-inducing dependence of the data representation from protected factors; and b) training set alteration via intelligent augmentation to address bias-causing data imbalance, by using generative models that allow the fine control of sensitive factors related to underrepresented populations via domain adaptation and latent space manipulation. When testing our methods on image analytics, experiments demonstrate that TARA significantly or fully debiases baseline models while outperforming competing debiasing methods that have the same amount of information, e.g., with (% overall accuracy, % accuracy gap) = (78.8, 0.5) vs. the baseline method's score of (71.8, 10.5) for EyePACS, and (73.7, 11.8) vs. (69.1, 21.7) for CelebA. Furthermore, recognizing certain limitations in current metrics used for assessing debiasing performance, we propose novel conjunctive debiasing metrics. Our experiments also demonstrate the ability of these novel metrics in assessing the Pareto efficiency of the proposed methods.

How can we build more just machine learning systems? To answer this question, we need to know both what justice is and how to tell whether one system is more or less just than another. That is, we need both a definition and a measure of justice. Theories of distributive justice hold that justice can be measured (in part) in terms of the fair distribution of benefits and burdens across people in society. Recently, the field known as fair machine learning has turned to John Rawls's theory of distributive justice for inspiration and operationalization. However, philosophers known as capability theorists have long argued that Rawls's theory uses the wrong measure of justice, thereby encoding biases against people with disabilities. If these theorists are right, is it possible to operationalize Rawls's theory in machine learning systems without also encoding its biases? In this paper, I draw on examples from fair machine learning to suggest that the answer to this question is no: the capability theorists' arguments against Rawls's theory carry over into machine learning systems. But capability theorists don't only argue that Rawls's theory uses the wrong measure, they also offer an alternative measure. Which measure of justice is right? And has fair machine learning been using the wrong one?

Artificial Intelligence Fairness in the Context of Accessibility Research on Intelligent Systems for People who are Deaf or Hard of Hearing

Sushant Kafle, Abraham Glasser, Sedeeq Al-khazraji +3

We discuss issues of Artificial Intelligence (AI) fairness for people with disabilities, with examples drawn from our research on human-computer interaction (HCI) for AI-based systems for people who are Deaf or Hard of Hearing (DHH). In particular, we discuss the need for inclusion of data from people with disabilities in training sets, the lack of interpretability of AI systems, ethical responsibilities of access technology researchers and companies, the need for appropriate evaluation metrics for AI-based access technologies (to determine if they are ready to be deployed and if they can be trusted by users), and the ways in which AI systems influence human behavior and influence the set of abilities needed by users to successfully interact with computing systems.

Machine learning (ML) models are increasingly used for personnel assessment and selection (e.g., resume screeners, automatically scored interviews). However, concerns have been raised throughout society that ML assessments may be biased and perpetuate or exacerbate inequality. Although organizational researchers have begun investigating ML assessments from traditional psychometric and legal perspectives, there is a need to understand, clarify, and integrate fairness operationalizations and algorithmic bias mitigation methods from the computer science, data science, and organizational research literatures. We present a four-stage model of developing ML assessments and applying bias mitigation methods, including 1) generating the training data, 2) training the model, 3) testing the model, and 4) deploying the model. When introducing the four-stage model, we describe potential sources of bias and unfairness at each stage. Then, we systematically review definitions and operationalizations of algorithmic bias, legal requirements governing personnel selection from the United States and Europe, and research on algorithmic bias mitigation across multiple domains and integrate these findings into our framework. Our review provides insights for both research and practice by elucidating possible mechanisms of algorithmic bias while identifying which bias mitigation methods are legal and effective. This integrative framework also reveals gaps in the knowledge of algorithmic bias mitigation that should be addressed by future collaborative research between organizational researchers, computer scientists, and data scientists. We provide recommendations for developing and deploying ML assessments, as well as recommendations for future research into algorithmic bias and fairness.

Artificial Intelligence is currently and rapidly changing the way organizations and businesses operate. Ethical leadership has become significantly important since organizations and businesses across various sectors are evolving with AI. Organizations and businesses may be facing several challenges and potential opportunities when using AI. Ethical leadership plays a central role in guiding organizations in facing those challenges and maximizing on those opportunities. This article explores the essence of ethical leadership in the age of AI, starting with a simplified introduction of ethical leadership and AI, then dives into an understanding of ethical leadership, its characteristics and importance, the ethical challenges AI causes including bias in AI algorithms. The opportunities for ethical leadership in the age of AI answers the question: What actionable strategies can leaders employ to address the challenges and leverage opportunities? and describes the benefits for organizations through these opportunities. A proposed framework for ethical leadership is presented in this article, incorporating the core components: fairness, transparency, sustainability etc. Through the importance of interdisciplinary collaboration, case studies of ethical leadership in AI, and recommendations, this article emphasizes that ethical leadership in the age of AI is morally essential and strategically advantageous.

The rapid advancement of foundation models in medical imaging represents a significant leap toward enhancing diagnostic accuracy and personalized treatment. However, the deployment of foundation models in healthcare necessitates a rigorous examination of their trustworthiness, encompassing privacy, robustness, reliability, explainability, and fairness. The current body of survey literature on foundation models in medical imaging reveals considerable gaps, particularly in the area of trustworthiness. Additionally, existing surveys on the trustworthiness of foundation models do not adequately address their specific variations and applications within the medical imaging domain. This survey aims to fill that gap by presenting a novel taxonomy of foundation models used in medical imaging and analyzing the key motivations for ensuring their trustworthiness. We review current research on foundation models in major medical imaging applications, focusing on segmentation, medical report generation, medical question and answering (Q\&A), and disease diagnosis. These areas are highlighted because they have seen a relatively mature and substantial number of foundation models compared to other applications. We focus on literature that discusses trustworthiness in medical image analysis manuscripts. We explore the complex challenges of building trustworthy foundation models for each application, summarizing current concerns and strategies for enhancing trustworthiness. Furthermore, we examine the potential of these models to revolutionize patient care. Our analysis underscores the imperative for advancing towards trustworthy AI in medical image analysis, advocating for a balanced approach that fosters innovation while ensuring ethical and equitable healthcare delivery.

This paper presents a critical, qualitative study of the social role of algorithmic bias in the context of the Chicago crime prediction algorithm, a predictive policing tool that forecasts when and where in the city crime is most likely to occur. Through interviews with 18 Chicago-area community organizations, academic researchers, and public sector actors, we show that stakeholders from different groups articulate diverse problem diagnoses of the tool's algorithmic bias, strategically using it as evidence to advance criminal justice interventions that align with stakeholders' positionality and political ends. Drawing inspiration from Catherine D'Ignazio's taxonomy of "refusing and using" data, we find that stakeholders use evidence of algorithmic bias to reform the policies around police patrol allocation; reject algorithm-based policing interventions; reframe crime as a structural rather than interpersonal problem; reveal data on authority figures in an effort to subvert their power; repair and heal families and communities; and, in the case of more powerful actors, to reaffirm their own authority or existing power structures. We identify the implicit assumptions and scope of these varied uses of algorithmic bias as evidence, showing that they require different (and sometimes conflicting) values about policing and AI. This divergence reflects long-standing tensions in the criminal justice reform landscape between the values of liberation and healing often centered by system-impacted communities and the values of surveillance and deterrence often instantiated in data-driven reform measures. We advocate for centering the interests and experiential knowledge of communities impacted by incarceration to ensure that evidence of algorithmic bias can serve as a device to challenge the status quo.

Collect, Measure, Repeat: Reliability Factors for Responsible AI Data Collection

Oana Inel, Tim Draws, Lora Aroyo

The rapid entry of machine learning approaches in our daily activities and high-stakes domains demands transparency and scrutiny of their fairness and reliability. To help gauge machine learning models' robustness, research typically focuses on the massive datasets used for their deployment, e.g., creating and maintaining documentation for understanding their origin, process of development, and ethical considerations. However, data collection for AI is still typically a one-off practice, and oftentimes datasets collected for a certain purpose or application are reused for a different problem. Additionally, dataset annotations may not be representative over time, contain ambiguous or erroneous annotations, or be unable to generalize across issues or domains. Recent research has shown these practices might lead to unfair, biased, or inaccurate outcomes. We argue that data collection for AI should be performed in a responsible manner where the quality of the data is thoroughly scrutinized and measured through a systematic set of appropriate metrics. In this paper, we propose a Responsible AI (RAI) methodology designed to guide the data collection with a set of metrics for an iterative in-depth analysis of the factors influencing the quality and reliability} of the generated data. We propose a granular set of measurements to inform on the internal reliability of a dataset and its external stability over time. We validate our approach across nine existing datasets and annotation tasks and four content modalities. This approach impacts the assessment of data robustness used for AI applied in the real world, where diversity of users and content is eminent. Furthermore, it deals with fairness and accountability aspects in data collection by providing systematic and transparent quality analysis for data collections.