Research Papers

Many internet applications are powered by machine learned models, which are usually trained on labeled datasets obtained through either implicit / explicit user feedback signals or human judgments. Since societal biases may be present in the generation of such datasets, it is possible for the trained models to be biased, thereby resulting in potential discrimination and harms for disadvantaged groups. Motivated by the need for understanding and addressing algorithmic bias in web-scale ML systems and the limitations of existing fairness toolkits, we present the LinkedIn Fairness Toolkit (LiFT), a framework for scalable computation of fairness metrics as part of large ML systems. We highlight the key requirements in deployed settings, and present the design of our fairness measurement system. We discuss the challenges encountered in incorporating fairness tools in practice and the lessons learned during deployment at LinkedIn. Finally, we provide open problems based on practical experience.

The rapid progress in Large Language Models (LLMs) could transform many fields, but their fast development creates significant challenges for oversight, ethical creation, and building user trust. This comprehensive review looks at key trust issues in LLMs, such as unintended harms, lack of transparency, vulnerability to attacks, alignment with human values, and environmental impact. Many obstacles can undermine user trust, including societal biases, opaque decision-making, potential for misuse, and the challenges of rapidly evolving technology. Addressing these trust gaps is critical as LLMs become more common in sensitive areas like finance, healthcare, education, and policy. To tackle these issues, we suggest combining ethical oversight, industry accountability, regulation, and public involvement. AI development norms should be reshaped, incentives aligned, and ethics integrated throughout the machine learning process, which requires close collaboration across technology, ethics, law, policy, and other fields. Our review contributes a robust framework to assess trust in LLMs and analyzes the complex trust dynamics in depth. We provide contextualized guidelines and standards for responsibly developing and deploying these powerful AI systems. This review identifies key limitations and challenges in creating trustworthy AI. By addressing these issues, we aim to build a transparent, accountable AI ecosystem that benefits society while minimizing risks. Our findings provide valuable guidance for researchers, policymakers, and industry leaders striving to establish trust in LLMs and ensure they are used responsibly across various applications for the good of society.

ConvXAI: Delivering Heterogeneous AI Explanations via Conversations to Support Human-AI Scientific Writing

Hua Shen, Chieh-Yang Huang, Tongshuang Wu +1

Despite a surge collection of XAI methods, users still struggle to obtain required AI explanations. Previous research suggests chatbots as dynamic solutions, but the effective design of conversational XAI agents for practical human needs remains under-explored. This paper focuses on Conversational XAI for AI-assisted scientific writing tasks. Drawing from human linguistic theories and formative studies, we identify four design rationales: "multifaceted", "controllability", "mix-initiative", "context-aware drill-down". We incorporate them into an interactive prototype, ConvXAI, which facilitates heterogeneous AI explanations for scientific writing through dialogue. In two studies with 21 users, ConvXAI outperforms a GUI-based baseline on improving human-perceived understanding and writing improvement. The paper further discusses the practical human usage patterns in interacting with ConvXAI for scientific co-writing.

Data-driven artificial intelligence models require explainability in intelligent manufacturing to streamline adoption and trust in modern industry. However, recently developed explainable artificial intelligence (XAI) techniques that estimate feature contributions on a model-agnostic level such as SHapley Additive exPlanations (SHAP) have not yet been evaluated for semi-supervised fault diagnosis and prognosis problems characterized by class imbalance and weakly labeled datasets. This paper explores the potential of utilizing Shapley values for a new clustering framework compatible with semi-supervised learning problems, loosening the strict supervision requirement of current XAI techniques. This broad methodology is validated on two case studies: a heatmap image dataset obtained from a semiconductor manufacturing process featuring class imbalance, and a benchmark dataset utilized in the 2021 Prognostics and Health Management (PHM) Data Challenge. Semi-supervised clustering based on Shapley values significantly improves upon clustering quality compared to the fully unsupervised case, deriving information-dense and meaningful clusters that relate to underlying fault diagnosis model predictions. These clusters can also be characterized by high-precision decision rules in terms of original feature values, as demonstrated in the second case study. The rules, limited to 1-2 terms utilizing original feature scales, describe 12 out of the 16 derived equipment failure clusters with precision exceeding 0.85, showcasing the promising utility of the explainable clustering framework for intelligent manufacturing applications.

Despite the popularity of Vision Transformers (ViTs) and eXplainable AI (XAI), only a few explanation methods have been designed specially for ViTs thus far. They mostly use attention weights of the [CLS] token on patch embeddings and often produce unsatisfactory saliency maps. This paper proposes a novel method for explaining ViTs called ViT-CX. It is based on patch embeddings, rather than attentions paid to them, and their causal impacts on the model output. Other characteristics of ViTs such as causal overdetermination are also considered in the design of ViT-CX. The empirical results show that ViT-CX produces more meaningful saliency maps and does a better job revealing all important evidence for the predictions than previous methods. The explanation generated by ViT-CX also shows significantly better faithfulness to the model. The codes and appendix are available at https://github.com/vaynexie/CausalX-ViT.

Data-driven predictive algorithms are widely used to automate and guide high-stake decision making such as bail and parole recommendation, medical resource distribution, and mortgage allocation. Nevertheless, harmful outcomes biased against vulnerable groups have been reported. The growing research field known as 'algorithmic fairness' aims to mitigate these harmful biases. Its primary methodology consists in proposing mathematical metrics to address the social harms resulting from an algorithm's biased outputs. The metrics are typically motivated by -- or substantively rooted in -- ideals of distributive justice, as formulated by political and legal philosophers. The perspectives of feminist political philosophers on social justice, by contrast, have been largely neglected. Some feminist philosophers have criticized the paradigm of distributive justice and have proposed corrective amendments to surmount its limitations. The present paper brings some key insights of feminist political philosophy to algorithmic fairness. The paper has three goals. First, I show that algorithmic fairness does not accommodate structural injustices in its current scope. Second, I defend the relevance of structural injustices -- as pioneered in the contemporary philosophical literature by Iris Marion Young -- to algorithmic fairness. Third, I take some steps in developing the paradigm of 'responsible algorithmic fairness' to correct for errors in the current scope and implementation of algorithmic fairness.

Discussions of algorithmic bias tend to focus on examples where either the data or the people building the algorithms are biased. This gives the impression that clean data and good intentions could eliminate bias. The neutrality of the algorithms themselves is defended by prominent Artificial Intelligence researchers. However, algorithms are not neutral. In addition to biased data and biased algorithm makers, AI algorithms themselves can be biased. This is illustrated with the example of collaborative filtering, which is known to suffer from popularity, and homogenizing biases. Iterative information filtering algorithms in general create a selection bias in the course of learning from user responses to documents that the algorithm recommended. These are not merely biases in the statistical sense; these statistical biases can cause discriminatory outcomes. Data points on the margins of distributions of human data tend to correspond to marginalized people. Popularity and homogenizing biases have the effect of further marginalizing the already marginal. This source of bias warrants serious attention given the ubiquity of algorithmic decision-making.

We introduce a variable importance measure to quantify the impact of individual input variables to a black box function. Our measure is based on the Shapley value from cooperative game theory. Many measures of variable importance operate by changing some predictor values with others held fixed, potentially creating unlikely or even logically impossible combinations. Our cohort Shapley measure uses only observed data points. Instead of changing the value of a predictor we include or exclude subjects similar to the target subject on that predictor to form a similarity cohort. Then we apply Shapley value to the cohort averages. We connect variable importance measures from explainable AI to function decompositions from global sensitivity analysis. We introduce a squared cohort Shapley value that splits previously studied Shapley effects over subjects, consistent with a Shapley axiom.

Conservative Agency via Attainable Utility Preservation

Alexander Matt Turner, Dylan Hadfield-Menell, Prasad Tadepalli

Reward functions are easy to misspecify; although designers can make corrections after observing mistakes, an agent pursuing a misspecified reward function can irreversibly change the state of its environment. If that change precludes optimization of the correctly specified reward function, then correction is futile. For example, a robotic factory assistant could break expensive equipment due to a reward misspecification; even if the designers immediately correct the reward function, the damage is done. To mitigate this risk, we introduce an approach that balances optimization of the primary reward function with preservation of the ability to optimize auxiliary reward functions. Surprisingly, even when the auxiliary reward functions are randomly generated and therefore uninformative about the correctly specified reward function, this approach induces conservative, effective behavior.

Many nations are promoting the green transition in the energy sector to attain neutral carbon emissions by 2050. Smart Grid 2.0 (SG2) is expected to explore data-driven analytics and enhance communication technologies to improve the efficiency and sustainability of distributed renewable energy systems. These features are beyond smart metering and electric surplus distribution in conventional smart grids. Given the high dependence on communication networks to connect distributed microgrids in SG2, potential cascading failures of connectivity can cause disruption to data synchronization to the remote control systems. This paper reviews security threats and defense tactics for three stakeholders: power grid operators, communication network providers, and consumers. Through the survey, we found that SG2's stakeholders are particularly vulnerable to substation attacks/vandalism, malware/ransomware threats, blockchain vulnerabilities and supply chain breakdowns. Furthermore, incorporating artificial intelligence (AI) into autonomous energy management in distributed energy resources of SG2 creates new challenges. Accordingly, adversarial samples and false data injection on electricity reading and measurement sensors at power plants can fool AI-powered control functions and cause messy error-checking operations in energy storage, wrong energy estimation in electric vehicle charging, and even fraudulent transactions in peer-to-peer energy trading models. Scalable blockchain-based models, physical unclonable function, interoperable security protocols, and trustworthy AI models designed for managing distributed microgrids in SG2 are typical promising protection models for future research.

What is the context which gave rise to the obligation to carry out a Fundamental Rights Impact Assessment (FRIA) in the AI Act? How has assessment of the impact on fundamental rights been framed by the EU legislator in the AI Act? What methodological criteria should be followed in developing the FRIA? These are the three main research questions that this article aims to address, through both legal analysis of the relevant provisions of the AI Act and discussion of various possible models for assessment of the impact of AI on fundamental rights. The overall objective of this article is to fill existing gaps in the theoretical and methodological elaboration of the FRIA, as outlined in the AI Act. In order to facilitate the future work of EU and national bodies and AI operators in placing this key tool for human-centric and trustworthy AI at the heart of the EU approach to AI design and development, this article outlines the main building blocks of a model template for the FRIA. While this proposal is consistent with the rationale and scope of the AI Act, it is also applicable beyond the cases listed in Article 27 and can serve as a blueprint for other national and international regulatory initiatives to ensure that AI is fully consistent with human rights.

AI Risk Categorization Decoded (AIR 2024): From Government Regulations to Corporate Policies

Yi Zeng, Kevin Klyman, Andy Zhou +6

We present a comprehensive AI risk taxonomy derived from eight government policies from the European Union, United States, and China and 16 company policies worldwide, making a significant step towards establishing a unified language for generative AI safety evaluation. We identify 314 unique risk categories organized into a four-tiered taxonomy. At the highest level, this taxonomy encompasses System & Operational Risks, Content Safety Risks, Societal Risks, and Legal & Rights Risks. The taxonomy establishes connections between various descriptions and approaches to risk, highlighting the overlaps and discrepancies between public and private sector conceptions of risk. By providing this unified framework, we aim to advance AI safety through information sharing across sectors and the promotion of best practices in risk mitigation for generative AI models and systems.