The need for fair AI is increasingly clear in the era of general-purpose systems such as ChatGPT, Gemini, and other large language models (LLMs). However, the increasing complexity of human-AI interaction and its social impacts have raised questions of how fairness standards could be applied. Here, we review the technical frameworks that machine learning researchers have used to evaluate fairness, such as group fairness and fair representations, and find that their application to LLMs faces inherent limitations. We show that each framework either does not logically extend to LLMs or presents a notion of fairness that is intractable for LLMs, primarily due to the multitudes of populations affected, sensitive attributes, and use cases. To address these challenges, we develop guidelines for the more realistic goal of achieving fairness in particular use cases: the criticality of context, the responsibility of LLM developers, and the need for stakeholder participation in an iterative process of design and evaluation. Moreover, it may eventually be possible and even necessary to use the general-purpose capabilities of AI systems to address fairness challenges as a form of scalable AI-assisted alignment.

When OpenAI's new voice assistant said it was "doing fantastic" in a launch demo this month, Scarlett Johansson was not. The Hollywood star said she was "shocked, angered and in disbelief" that the updated version of ChatGPT, which can listen to spoken prompts and respond verbally, had a voice "eerily similar" to hers. One of Johansson's signature roles was as the voice of a futuristic version of Siri in the 2013 film Her and, for the actor, the similarity was stark. The OpenAI chief executive, Sam Altman, appeared to acknowledge the film's influence with a one-word post on X on the day of the launch: "her". In a statement, Johansson said Altman had approached her last year to be a voice of ChatGPT and that she had declined for "personal reasons".

This paper aims to advance our understanding of how Visual Language Models (VLMs) handle privacy-sensitive information, a crucial concern as these technologies become integral to everyday life. To this end, we introduce a new benchmark PrivBench, which contains images from 8 sensitive categories such as passports, or fingerprints. We evaluate 10 state-of-the-art VLMs on this benchmark and observe a generally limited understanding of privacy, highlighting a significant area for model improvement. Based on this we introduce PrivTune, a new instruction-tuning dataset aimed at equipping VLMs with knowledge about visual privacy. By tuning two pretrained VLMs, TinyLLaVa and MiniGPT-v2, on this small dataset, we achieve strong gains in their ability to recognize sensitive content, outperforming even GPT4-V. At the same time, we show that privacy-tuning only minimally affects the VLMs performance on standard benchmarks such as VQA. Overall, this paper lays out a crucial challenge for making VLMs effective in handling real-world data safely and provides a simple recipe that takes the first step towards building privacy-aware VLMs.

Dark patterns are deceptive user interfaces employed by e-commerce websites to manipulate user's behavior in a way that benefits the website, often unethically. This study investigates the detection of such dark patterns. Existing solutions include UIGuard, which uses computer vision and natural language processing, and approaches that categorize dark patterns based on detectability or utilize machine learning models trained on datasets. We propose combining web scraping techniques with fine-tuned BERT language models and generative capabilities to identify dark patterns, including outliers. The approach scrapes textual content, feeds it into the BERT model for detection, and leverages BERT's bidirectional analysis and generation abilities. The study builds upon research on automatically detecting and explaining dark patterns, aiming to raise awareness and protect consumers.

Fairness-aware Graph Neural Networks (GNNs) often face a challenging trade-off, where prioritizing fairness may require compromising utility. In this work, we re-examine fairness through the lens of spectral graph theory, aiming to reconcile fairness and utility within the framework of spectral graph learning. We explore the correlation between sensitive features and spectrum in GNNs, using theoretical analysis to delineate the similarity between original sensitive features and those after convolution under different spectrum. Our analysis reveals a reduction in the impact of similarity when the eigenvectors associated with the largest magnitude eigenvalue exhibit directional similarity. Based on these theoretical insights, we propose FUGNN, a novel spectral graph learning approach that harmonizes the conflict between fairness and utility. FUGNN ensures algorithmic fairness and utility by truncating the spectrum and optimizing eigenvector distribution during the encoding process. The fairness-aware eigenvector selection reduces the impact of convolution on sensitive features while concurrently minimizing the sacrifice of utility. FUGNN further optimizes the distribution of eigenvectors through a transformer architecture. By incorporating the optimized spectrum into the graph convolution network, FUGNN effectively learns node representations. Experiments on six real-world datasets demonstrate the superiority of FUGNN over baseline methods. The codes are available at https://github.com/yushuowiki/FUGNN.

Machine Learning and AI have the potential to transform data-driven scientific discovery, enabling accurate predictions for several scientific phenomena. As many scientific questions are inherently causal, this paper looks at the causal inference task of treatment effect estimation, where we assume binary effects that are recorded as high-dimensional images in a Randomized Controlled Trial (RCT). Despite being the simplest possible setting and a perfect fit for deep learning, we theoretically find that many common choices in the literature may lead to biased estimates. To test the practical impact of these considerations, we recorded the first real-world benchmark for causal inference downstream tasks on high-dimensional observations as an RCT studying how garden ants (Lasius neglectus) respond to microparticles applied onto their colony members by hygienic grooming. Comparing 6 480 models fine-tuned from state-of-the-art visual backbones, we find that the sampling and modeling choices significantly affect the accuracy of the causal estimate, and that classification accuracy is not a proxy thereof. We further validated the analysis, repeating it on a synthetically generated visual data set controlling the causal model. Our results suggest that future benchmarks should carefully consider real downstream scientific questions, especially causal ones. Further, we highlight guidelines for representation learning methods to help answer causal questions in the sciences. All code and data will be released.

Artificial Intelligence techniques such as Machine Learning (ML) have not been exploited to their maximum potential in the legal domain. This has been partially due to the insufficient explanations they provided about their decisions. Automatic expert systems with explanatory capabilities can be specially useful when legal practitioners search jurisprudence to gather contextual knowledge for their cases. Therefore, we propose a hybrid system that applies ML for multi-label classification of judgements (sentences) and visual and natural language descriptions for explanation purposes, boosted by Natural Language Processing techniques and deep legal reasoning to identify the entities, such as the parties, involved. We are not aware of any prior work on automatic multi-label classification of legal judgements also providing natural language explanations to the end-users with comparable overall quality. Our solution achieves over 85 % micro precision on a labelled data set annotated by legal experts. This endorses its interest to relieve human experts from monotonous labour-intensive legal classification tasks.

The rapid proliferation of generative AI has raised questions about the competitiveness of lower-parameter, locally tunable, open-weight models relative to high-parameter, API-guarded, closed-weight models in terms of performance, domain adaptation, cost, and generalization. Centering under-resourced yet risk-intolerant settings in government, research, and healthcare, we see for-profit closed-weight models as incompatible with requirements for transparency, privacy, adaptability, and standards of evidence. Yet the performance penalty in using open-weight models, especially in low-data and low-resource settings, is unclear. We assess the feasibility of using smaller, open-weight models to replace GPT-4-Turbo in zero-shot, few-shot, and fine-tuned regimes, assuming access to only a single, low-cost GPU. We assess value-sensitive issues around bias, privacy, and abstention on three additional tasks relevant to those topics. We find that with relatively low effort, very low absolute monetary cost, and relatively little data for fine-tuning, small open-weight models can achieve competitive performance in domain-adapted tasks without sacrificing generality. We then run experiments considering practical issues in bias, privacy, and hallucination risk, finding that open models offer several benefits over closed models. We intend this work as a case study in understanding the opportunity cost of reproducibility and transparency over for-profit state-of-the-art zero shot performance, finding this cost to be marginal under realistic settings.

In this paper, we propose a new annotation scheme to classify different types of clauses in Terms-and-Conditions contracts with the ultimate goal of supporting legal experts to quickly identify and assess problematic issues in this type of legal documents. To this end, we built a small corpus of Terms-and-Conditions contracts and finalized an annotation scheme of 14 categories, eventually reaching an inter-annotator agreement of 0.92. Then, for 11 of them, we experimented with binary classification tasks using few-shot prompting with a multilingual T5 and two fine-tuned versions of two BERT-based LLMs for Italian. Our experiments showed the feasibility of automatic classification of our categories by reaching accuracies ranging from .79 to .95 on validation tasks.

The reliance of language model training on massive amounts of computation and vast datasets scraped from potentially low-quality, copyrighted, or sensitive data has come into question practically, legally, and ethically. Federated learning provides a plausible alternative by enabling previously untapped data to be voluntarily gathered from collaborating organizations. However, when scaled globally, federated learning requires collaboration across heterogeneous legal, security, and privacy regimes while accounting for the inherent locality of language data; this further exacerbates the established challenge of federated statistical heterogeneity. We propose a Worldwide Federated Language Model Training~(WorldLM) system based on federations of federations, where each federation has the autonomy to account for factors such as its industry, operating jurisdiction, or competitive environment. WorldLM enables such autonomy in the presence of statistical heterogeneity via partial model localization by allowing sub-federations to attentively aggregate key layers from their constituents. Furthermore, it can adaptively share information across federations via residual layer embeddings. Evaluations of language modeling on naturally heterogeneous datasets show that WorldLM outperforms standard federations by up to $1.91\times$, approaches the personalized performance of fully local models, and maintains these advantages under privacy-enhancing techniques.