The issue of group fairness in machine learning models, where certain sub-populations or groups are favored over others, has been recognized for some time. While many mitigation strategies have been proposed in centralized learning, many of these methods are not directly applicable in federated learning, where data is privately stored on multiple clients. To address this, many proposals try to mitigate bias at the level of clients before aggregation, which we call locally fair training. However, the effectiveness of these approaches is not well understood. In this work, we investigate the theoretical foundation of locally fair training by studying the relationship between global model fairness and local model fairness. Additionally, we prove that for a broad class of fairness metrics, the global model's fairness can be obtained using only summary statistics from local clients. Based on that, we propose a globally fair training algorithm that directly minimizes the penalized empirical loss. Real-data experiments demonstrate the promising performance of our proposed approach for enhancing fairness while retaining high accuracy compared to locally fair training methods.

In this paper, we trace the biases in current natural language processing (NLP) models back to their origins in racism, sexism, and homophobia over the last 500 years. We review literature from critical race theory, gender studies, data ethics, and digital humanities studies, and summarize the origins of bias in NLP models from these social science perspective. We show how the causes of the biases in the NLP pipeline are rooted in social issues. Finally, we argue that the only way to fix the bias and unfairness in NLP is by addressing the social problems that caused them in the first place and by incorporating social sciences and social scientists in efforts to mitigate bias in NLP models. We provide actionable recommendations for the NLP research community to do so.

Foreign allies and adversaries alike have pushed AI regulations, but Congress has stalled. Lawmakers told Fox News bipartisan efforts are needed to regulate the space. WASHINGTON, D.C. – Members of Congress provided a range of opinions on regulating AI, but several agreed that bipartisanship is the key to moving forward with a framework, lawmakers on Capitol Hill told Fox News. China and the European Union have recently drafted AI regulations, but Congress hasn't passed any legislation since the tech's recent rapid development. Republicans worry that lawmakers could overregulate AI and harm innovation, while Democrats fear that machine learning poses potential threats to consumers.

Earlier this month, Lina Khan, chair of the US Federal Trade Commission (FTC), wrote an essay in The New York Times affirming the agency's commitment to regulating AI. But there was one AI application Khan didn't mention that the FTC urgently needs to regulate: automated hiring systems. These range in complexity from tools that merely parse resumes and rank them to systems that green-light candidates and trash applicants deemed unfit. Increasingly, working Americans are obligated to use them if they want to get hired. If you buy something using links in our stories, we may earn a commission.

The EU has approved Microsoft's $69bn (£55bn) acquisition of the Call of Duty creator Activision Blizzard, in a move that puts Brussels at loggerheads with its UK counterpart over the gaming mega-deal. The EU accepted Microsoft's concessions on cloud gaming, the same problem that led the Competition and Markets Authority to block the transaction last month. The proposed deal would bring together Microsoft, the maker of the Xbox console, with the video game developer behind titles including World of Warcraft, Hearthstone, Candy Crush Saga and Overwatch. The move by the European Commission, the bloc's executive arm, will revive Microsoft's hopes for the deal as it prepares to appeal against the CMA's decision. The Federal Trade Commission in the US has also come out against the takeover and is suing to block it.

Hosted by Ben Byford, The Machine Ethics Podcast brings together interviews with academics, authors, business leaders, designers and engineers on the subject of autonomous algorithms, artificial intelligence, machine learning, and technology's impact on society. This episode we're talking with Dr Marie Oldfield on definitions of AI, the education and communication gaps with AI, explainable models, ethics in education, problems with audits and legislation, AI accreditation, importance of interdisciplinary teams, when to use AI or not, and harms from algorithms. Marie Oldfield (CStat, CSci, FIScT) is the CEO of Oldfield Consultancy and Kuinua Coaching. She is also a Senior Lecturer in Practice for the London School of Economics. With a background in mathematics and philosophy, she is a trusted advisor to government, defence, and the legal sector, amongst others.

These days, so-called generative AI can (allegedly) make art, write books, and compose poetry. Systems like Stable Diffusion, Midjourney, and ChatGPT are seemingly quite good at it. But for some artists, this creates problems. Namely, determining what legal rights they have when their work is scraped by these tools. Faced by the rise in these systems, authors and artists are pushing back.

NLP in the legal domain has seen increasing success with the emergence of Transformer-based Pre-trained Language Models (PLMs) pre-trained on legal text. PLMs trained over European and US legal text are available publicly; however, legal text from other domains (countries), such as India, have a lot of distinguishing characteristics. With the rapidly increasing volume of Legal NLP applications in various countries, it has become necessary to pre-train such LMs over legal text of other countries as well. In this work, we attempt to investigate pre-training in the Indian legal domain. We re-train (continue pre-training) two popular legal PLMs, LegalBERT and CaseLawBERT, on Indian legal data, as well as train a model from scratch with a vocabulary based on Indian legal text. We apply these PLMs over three benchmark legal NLP tasks -- Legal Statute Identification from facts, Semantic Segmentation of Court Judgment Documents, and Court Appeal Judgment Prediction -- over both Indian and non-Indian (EU, UK) datasets. We observe that our approach not only enhances performance on the new domain (Indian texts) but also over the original domain (European and UK texts). We also conduct explainability experiments for a qualitative comparison of all these different PLMs.

This paper tackles the task of legal extractive summarization using a dataset of 430K U.S. court opinions with key passages annotated. According to automated summary quality metrics, the reinforcement-learning-based MemSum model is best and even out-performs transformer-based models. In turn, expert human evaluation shows that MemSum summaries effectively capture the key points of lengthy court opinions. Motivated by these results, we open-source our models to the general public. This represents progress towards democratizing law and making U.S. court opinions more accessible to the general public.

The development of large-scale Chinese language models is flourishing, yet there is a lack of corresponding capability assessments. Therefore, we propose a test to measure the multitask accuracy of large Chinese language models. This test encompasses four major domains, including medicine, law, psychology, and education, with 15 subtasks in medicine and 8 subtasks in education. We found that the best-performing models in the zero-shot setting outperformed the worst-performing models by nearly 18.6 percentage points on average. Across the four major domains, the highest average zero-shot accuracy of all models is 0.512. In the subdomains, only the GPT-3.5-turbo model achieved a zero-shot accuracy of 0.693 in clinical medicine, which was the highest accuracy among all models across all subtasks. All models performed poorly in the legal domain, with the highest zero-shot accuracy reaching only 0.239. By comprehensively evaluating the breadth and depth of knowledge across multiple disciplines, this test can more accurately identify the shortcomings of the models.