Research Papers

This survey article assesses and compares existing critiques of current fairness-enhancing technical interventions into machine learning (ML) that draw from a range of non-computing disciplines, including philosophy, feminist studies, critical race and ethnic studies, legal studies, anthropology, and science and technology studies. It bridges epistemic divides in order to offer an interdisciplinary understanding of the possibilities and limits of hegemonic computational approaches to ML fairness for producing just outcomes for society's most marginalized. The article is organized according to nine major themes of critique wherein these different fields intersect: 1) how "fairness" in AI fairness research gets defined; 2) how problems for AI systems to address get formulated; 3) the impacts of abstraction on how AI tools function and its propensity to lead to technological solutionism; 4) how racial classification operates within AI fairness research; 5) the use of AI fairness measures to avoid regulation and engage in ethics washing; 6) an absence of participatory design and democratic deliberation in AI fairness considerations; 7) data collection practices that entrench "bias," are non-consensual, and lack transparency; 8) the predatory inclusion of marginalized groups into AI systems; and 9) a lack of engagement with AI's long-term social and ethical outcomes. Drawing from these critiques, the article concludes by imagining future ML fairness research directions that actively disrupt entrenched power dynamics and structural injustices in society.

The paper presents a novel deep learning approach, which extracts latent information from trained Deep Neural Networks (DNNs) and derives concise representations that are analyzed in an effective, unified way for prediction purposes. It is well known that DNNs are capable of analyzing complex data; however, they lack transparency in their decision making, in the sense that it is not straightforward to justify their prediction, or to visualize the features on which the decision was based. Moreover, they generally require large amounts of data in order to learn and become able to adapt to different environments. This makes their use difficult in healthcare, where trust and personalization are key issues. Transparency combined with high prediction accuracy are the targeted goals of the proposed approach. It includes both supervised DNN training and unsupervised learning of latent variables extracted from the trained DNNs. Domain Adaptation from multiple sources is also presented as an extension, where the extracted latent variable representations are used to generate predictions in other, non-annotated, environments. Successful application is illustrated through a large experimental study in various fields: prediction of Parkinson's disease from MRI and DaTScans; prediction of COVID-19 and pneumonia from CT scans and X-rays; optical character verification in retail food packaging.

International AI Safety Report

Yoshua Bengio, Sören Mindermann, Daniel Privitera +93

The first International AI Safety Report comprehensively synthesizes the current evidence on the capabilities, risks, and safety of advanced AI systems. The report was mandated by the nations attending the AI Safety Summit in Bletchley, UK. Thirty nations, the UN, the OECD, and the EU each nominated a representative to the report's Expert Advisory Panel. A total of 100 AI experts contributed, representing diverse perspectives and disciplines. Led by the report's Chair, these independent experts collectively had full discretion over the report's content.

This study examines the extent to which U.S. federal agencies responded to and implemented the principles outlined in the White House's October 2022 "Blueprint for an AI Bill of Rights." The Blueprint provided a framework for the ethical governance of artificial intelligence systems, organized around five core principles: safety and effectiveness, protection against algorithmic discrimination, data privacy, notice and explanation about AI systems, and human alternatives and fallback. Through an analysis of publicly available records across 15 federal departments, the authors found limited evidence that the Blueprint directly influenced agency actions after its release. Only five departments explicitly mentioned the Blueprint, while 12 took steps aligned with one or more of its principles. However, much of this work appeared to have precedents predating the Blueprint or motivations disconnected from it, such as compliance with prior executive orders on trustworthy AI. Departments' activities often emphasized priorities like safety, accountability and transparency that overlapped with Blueprint principles, but did not necessarily stem from it. The authors conclude that the non-binding Blueprint seems to have had minimal impact on shaping the U.S. government's approach to ethical AI governance in its first year. Factors like public concerns after high-profile AI releases and obligations to follow direct executive orders likely carried more influence over federal agencies. More rigorous study would be needed to definitively assess the Blueprint's effects within the federal bureaucracy and broader society.

In the development of governmental policy for artificial intelligence (AI) that is informed by ethics, one avenue currently pursued is that of drawing on AI Ethics Principles. However, these AI Ethics Principles often fail to be actioned in governmental policy. This paper proposes a novel framework for the development of Actionable Principles for AI. The approach acknowledges the relevance of AI Ethics Principles and homes in on methodological elements to increase their practical implementability in policy processes. As a case study, elements are extracted from the development process of the Ethics Guidelines for Trustworthy AI of the European Commissions High Level Expert Group on AI. Subsequently, these elements are expanded on and evaluated in light of their ability to contribute to a prototype framework for the development of Actionable Principles for AI. The paper proposes the following three propositions for the formation of such a prototype framework: (1) preliminary landscape assessments; (2) multi-stakeholder participation and cross-sectoral feedback; and, (3) mechanisms to support implementation and operationalizability.

The 'robot rights' debate, and its related question of 'robot responsibility', invokes some of the most polarized positions in AI ethics. While some advocate for granting robots rights on a par with human beings, others, in a stark opposition argue that robots are not deserving of rights but are objects that should be our slaves. Grounded in post-Cartesian philosophical foundations, we argue not just to deny robots 'rights', but to deny that robots, as artifacts emerging out of and mediating human being, are the kinds of things that could be granted rights in the first place. Once we see robots as mediators of human being, we can understand how the `robots rights' debate is focused on first world problems, at the expense of urgent ethical concerns, such as machine bias, machine elicited human labour exploitation, and erosion of privacy all impacting society's least privileged individuals. We conclude that, if human being is our starting point and human welfare is the primary concern, the negative impacts emerging from machinic systems, as well as the lack of taking responsibility by people designing, selling and deploying such machines, remains the most pressing ethical discussion in AI.

How Case Based Reasoning Explained Neural Networks: An XAI Survey of Post-Hoc Explanation-by-Example in ANN-CBR Twins

Mark T Keane, Eoin M Kenny

This paper surveys an approach to the XAI problem, using post-hoc explanation by example, that hinges on twinning Artificial Neural Networks (ANNs) with Case-Based Reasoning (CBR) systems, so-called ANN-CBR twins. A systematic survey of 1100+ papers was carried out to identify the fragmented literature on this topic and to trace it influence through to more recent work involving Deep Neural Networks (DNNs). The paper argues that this twin-system approach, especially using ANN-CBR twins, presents one possible coherent, generic solution to the XAI problem (and, indeed, XCBR problem). The paper concludes by road-mapping some future directions for this XAI solution involving (i) further tests of feature-weighting techniques, (iii) explorations of how explanatory cases might best be deployed (e.g., in counterfactuals, near-miss cases, a fortori cases), and (iii) the raising of the unwelcome and, much ignored, issue of human user evaluation.

XAIR: A Framework of Explainable AI in Augmented Reality

Xuhai Xu, Mengjie Yu, Tanya R. Jonker +15

Explainable AI (XAI) has established itself as an important component of AI-driven interactive systems. With Augmented Reality (AR) becoming more integrated in daily lives, the role of XAI also becomes essential in AR because end-users will frequently interact with intelligent services. However, it is unclear how to design effective XAI experiences for AR. We propose XAIR, a design framework that addresses "when", "what", and "how" to provide explanations of AI output in AR. The framework was based on a multi-disciplinary literature review of XAI and HCI research, a large-scale survey probing 500+ end-users' preferences for AR-based explanations, and three workshops with 12 experts collecting their insights about XAI design in AR. XAIR's utility and effectiveness was verified via a study with 10 designers and another study with 12 end-users. XAIR can provide guidelines for designers, inspiring them to identify new design opportunities and achieve effective XAI designs in AR.

Human Perceptions on Moral Responsibility of AI: A Case Study in AI-Assisted Bail Decision-Making

Gabriel Lima, Nina Grgić-Hlača, Meeyoung Cha

How to attribute responsibility for autonomous artificial intelligence (AI) systems' actions has been widely debated across the humanities and social science disciplines. This work presents two experiments ($N$=200 each) that measure people's perceptions of eight different notions of moral responsibility concerning AI and human agents in the context of bail decision-making. Using real-life adapted vignettes, our experiments show that AI agents are held causally responsible and blamed similarly to human agents for an identical task. However, there was a meaningful difference in how people perceived these agents' moral responsibility; human agents were ascribed to a higher degree of present-looking and forward-looking notions of responsibility than AI agents. We also found that people expect both AI and human decision-makers and advisors to justify their decisions regardless of their nature. We discuss policy and HCI implications of these findings, such as the need for explainable AI in high-stakes scenarios.

The diffusion of artificial intelligence (AI) applications in organizations and society has fueled research on explaining AI decisions. The explainable AI (xAI) field is rapidly expanding with numerous ways of extracting information and visualizing the output of AI technologies (e.g. deep neural networks). Yet, we have a limited understanding of how xAI research addresses the need for explainable AI. We conduct a systematic review of xAI literature on the topic and identify four thematic debates central to how xAI addresses the black-box problem. Based on this critical analysis of the xAI scholarship we synthesize the findings into a future research agenda to further the xAI body of knowledge.

Explainable Goal-Driven Agents and Robots -- A Comprehensive Review

Fatai Sado, Chu Kiong Loo, Wei Shiung Liew +2

Recent applications of autonomous agents and robots, such as self-driving cars, scenario-based trainers, exploration robots, and service robots have brought attention to crucial trust-related challenges associated with the current generation of artificial intelligence (AI) systems. AI systems based on the connectionist deep learning neural network approach lack capabilities of explaining their decisions and actions to others, despite their great successes. Without symbolic interpretation capabilities, they are black boxes, which renders their decisions or actions opaque, making it difficult to trust them in safety-critical applications. The recent stance on the explainability of AI systems has witnessed several approaches on eXplainable Artificial Intelligence (XAI); however, most of the studies have focused on data-driven XAI systems applied in computational sciences. Studies addressing the increasingly pervasive goal-driven agents and robots are still missing. This paper reviews approaches on explainable goal-driven intelligent agents and robots, focusing on techniques for explaining and communicating agents perceptual functions (example, senses, and vision) and cognitive reasoning (example, beliefs, desires, intention, plans, and goals) with humans in the loop. The review highlights key strategies that emphasize transparency, understandability, and continual learning for explainability. Finally, the paper presents requirements for explainability and suggests a roadmap for the possible realization of effective goal-driven explainable agents and robots.

On the Algorithmic Bias of Aligning Large Language Models with RLHF: Preference Collapse and Matching Regularization

Jiancong Xiao, Ziniu Li, Xingyu Xie +4

Accurately aligning large language models (LLMs) with human preferences is crucial for informing fair, economically sound, and statistically efficient decision-making processes. However, we argue that the predominant approach for aligning LLMs with human preferences through a reward model -- reinforcement learning from human feedback (RLHF) -- suffers from an inherent algorithmic bias due to its Kullback--Leibler-based regularization in optimization. In extreme cases, this bias could lead to a phenomenon we term preference collapse, where minority preferences are virtually disregarded. To mitigate this algorithmic bias, we introduce preference matching (PM) RLHF, a novel approach that provably aligns LLMs with the preference distribution of the reward model under the Bradley--Terry--Luce/Plackett--Luce model. Central to our approach is a PM regularizer that takes the form of the negative logarithm of the LLM's policy probability distribution over responses, which helps the LLM balance response diversification and reward maximization. Notably, we obtain this regularizer by solving an ordinary differential equation that is necessary for the PM property. For practical implementation, we introduce a conditional variant of PM RLHF that is tailored to natural language generation. Finally, we empirically validate the effectiveness of conditional PM RLHF through experiments on the OPT and Llama-family models, demonstrating a 29% to 41% improvement in alignment with human preferences, as measured by a certain metric, compared to standard RLHF.