A stereotype is a generalized perception of a specific group of humans. It is often potentially encoded in human language, which is more common in texts on social issues. Previous works simply define a sentence as stereotypical and anti-stereotypical. However, the stereotype of a sentence may require fine-grained quantification. In this paper, to fill this gap, we quantify stereotypes in language by annotating a dataset. We use the pre-trained language models (PLMs) to learn this dataset to predict stereotypes of sentences. Then, we discuss stereotypes about common social issues such as hate speech, sexism, sentiments, and disadvantaged and advantaged groups. We demonstrate the connections and differences between stereotypes and common social issues, and all four studies validate the general findings of the current studies. In addition, our work suggests that fine-grained stereotype scores are a highly relevant and competitive dimension for research on social issues.

Hypothesis testing is an indispensable tool to answer scientific questions in the context of clinical trials, bioinformatics, social sciences, etc. The data within such domains often involves sensitive and private information pertaining to individuals. Researchers often bear legal obligations to safeguard the privacy of such data. In this context, differential privacy (Dwork, 2006) has emerged as a compelling framework for ensuring privacy in statistical analyses with confidential data. Consequently, differentially private versions of numerous well-established hypothesis tests have been developed, although exclusively from a frequentist view point. This encompasses private adaptations of test of binomial proportions (Awan and Slavković, 2018), linear regression (Alabi and Vadhan, 2023), goodness of fit (Kwak et al., 2021), analysis of variance (Swanberg et al., 2019), high-dimensional normal means (Narayanan, 2022), to name a few. Differentially private versions of common non-parametric tests (Couch et al., 2019), permutation tests (Kim and Schrab, 2023), etc., have emerged as

Most existing Legal Judgment Prediction (LJP) models focus on discovering the legal triggers in the criminal fact description. However, in real-world scenarios, a professional judge not only needs to assimilate the law case experience that thrives on past sentenced legal judgments but also depends on the professional legal grounded reasoning that learned from professional legal knowledge. In this paper, we propose a LegalDuet model, which pretrains language models to learn a tailored embedding space for making legal judgments. It proposes a dual-view legal clue reasoning mechanism, which derives from two reasoning chains of judges: 1) Law Case Reasoning, which makes legal judgments according to the judgment experiences learned from analogy/confusing legal cases; 2) Legal Ground Reasoning, which lies in matching the legal clues between criminal cases and legal decisions. Our experiments show that LegalDuet achieves state-of-the-art performance on the CAIL2018 dataset and outperforms baselines with about 4% improvements on average. Our dual-view reasoning based pretraining can capture critical legal clues to learn a tailored embedding space to distinguish criminal cases. It reduces LegalDuet's uncertainty during prediction and brings pretraining advances to the confusing/low frequent charges. All codes are available at https://github.com/NEUIR/LegalDuet.

This chapter focuses on gender-related errors in machine translation (MT) in the context of low-resource languages. We begin by explaining what low-resource languages are, examining the inseparable social and computational factors that create such linguistic hierarchies. We demonstrate through a case study of our mother tongue Bengali, a global language spoken by almost 300 million people but still classified as low-resource, how gender is assumed and inferred in translations to and from the high(est)-resource English when no such information is provided in source texts. We discuss the postcolonial and societal impacts of such errors leading to linguistic erasure and representational harms, and conclude by discussing potential solutions towards uplifting languages by providing them more agency in MT conversations.

In response to recent data regulation requirements, machine unlearning (MU) has emerged as a critical process to remove the influence of specific examples from a given model. Although exact unlearning can be achieved through complete model retraining using the remaining dataset, the associated computational costs have driven the development of efficient, approximate unlearning techniques. Moving beyond data-centric MU approaches, our study introduces a novel model-based perspective: model sparsification via weight pruning, which is capable of reducing the gap between exact unlearning and approximate unlearning. We show in both theory and practice that model sparsity can boost the multi-criteria unlearning performance of an approximate unlearner, closing the approximation gap, while continuing to be efficient. This leads to a new MU paradigm, termed prune first, then unlearn, which infuses a sparse model prior into the unlearning process. Building on this insight, we also develop a sparsity-aware unlearning method that utilizes sparsity regularization to enhance the training process of approximate unlearning. Extensive experiments show that our proposals consistently benefit MU in various unlearning scenarios. A notable highlight is the 77% unlearning efficacy gain of fine-tuning (one of the simplest unlearning methods) when using sparsity-aware unlearning. Furthermore, we demonstrate the practical impact of our proposed MU methods in addressing other machine learning challenges, such as defending against backdoor attacks and enhancing transfer learning. Codes are available at https://github.com/OPTML-Group/Unlearn-Sparse.

Deepfake pornographic images of Taylor Swift have been shared across the social media platform X, highlighting the lack of digital privacy protections for victims across the globe. It isn't known who generated the fake images of Swift, which have been viewed tens of millions of times since Wednesday. On Friday, X said their team was working to remove all non-consensual nudity from their site, which is "strictly prohibited." "We're committed to maintaining a safe and respectful environment for all users," the company said. Swift has not publicly commented on the matter.

The estate of George Carlin is suing the media company behind a fake, hour-long comedy special whose creators boasted of using artificial intelligence to re-create the late standup comic's style and material. The lawsuit filed in federal court in Los Angeles on Thursday asks that a judge order the podcast outlet Dudesy to immediately take down the audio special, George Carlin: I'm Glad I'm Dead, in which a synthesis of Carlin delivers commentary on current events. Carlin's daughter, Kelly Carlin, said in a statement that the work is "a poorly-executed facsimile cobbled together by unscrupulous individuals to capitalize on the extraordinary goodwill my father established with his adoring fanbase". The named defendants are Dudesy and podcast hosts Will Sasso and Chad Kultgen. The defendants have not filed a response to the lawsuit and it was not clear whether they have retained an attorney.

If anyone can rally up a base, it's Taylor Swift. When sexually explicit, likely AI-generated images of Swift circulated on social media this week, it galvanized her fans. Swifties found phrases and hashtags related to the images and flooded them with videos and photos of Swift performing. "Protect Taylor Swift" went viral, trending as Swifties spoke out against not just the Swift deepfakes, but all nonconsensual, explicit images made of women. Swift, arguably the most famous woman in the world right now, has become the high-profile victim of an all-too-frequent form of harassment.

The rapid online spread of deepfake pornographic images of Taylor Swift has renewed calls, including from US politicians, to criminalise the practice, in which artificial intelligence is used to synthesise fake but convincing explicit imagery. The images of the US popstar have been distributed across social media and seen by millions this week. Previously distributed on the app Telegram, one of the images of Swift hosted on X was seen 47m times before it was removed. X said in a statement: "Our teams are actively removing all identified images and taking appropriate actions against the accounts responsible for posting them." Yvette D Clarke, a Democratic congresswoman for New York, wrote on X: "What's happened to Taylor Swift is nothing new. For yrs, women have been targets of deepfakes [without] their consent. And [with] advancements in AI, creating deepfakes is easier & cheaper. This is an issue both sides of the aisle & even Swifties should be able to come together to solve."

U.S. antitrust enforcers are opening an investigation into the relationships between leading artificial intelligence startups such as ChatGPT-maker OpenAI and Anthropic and the tech giants that have invested billions of dollars into them. "We're scrutinizing whether these ties enable dominant firms to exert undue influence or gain privileged access in ways that could undermine fair competition," said Lina Khan, chair of the U.S. Federal Trade Commission, in opening remarks at a Thursday AI forum. Khan said the market inquiry would review "the investments and partnerships being formed between AI developers and major cloud service providers." The FTC said on Thursday that it has issued "compulsory orders" to five companies -- cloud providers Amazon, Google and Microsoft, and AI startups Anthropic and OpenAI -- requiring them to provide information regarding investments and partnerships. Microsoft's close and years-long relationship with OpenAI is the best known of the partnerships.