Prompting Large Language Models (LLMs) has created new and interesting means for classifying textual data. While evaluating and remediating group fairness is a well-studied problem in classifier fairness literature, some classical approaches (e.g., regularization) do not carry over, and some new opportunities arise (e.g., prompt-based remediation). We measure fairness of LLM-based classifiers on a toxicity classification task, and empirically show that prompt-based classifiers may lead to unfair decisions. We introduce several remediation techniques and benchmark their fairness and performance trade-offs. We hope our work encourages more research on group fairness in LLM-based classifiers.

This report introduces a new family of multimodal models, Gemini, that exhibit remarkable capabilities across image, audio, video, and text understanding. The Gemini family consists of Ultra, Pro, and Nano sizes, suitable for applications ranging from complex reasoning tasks to on-device memory-constrained use-cases. Evaluation on a broad range of benchmarks shows that our most-capable Gemini Ultra model advances the state of the art in 30 of 32 of these benchmarks - notably being the first model to achieve human-expert performance on the well-studied exam benchmark MMLU, and improving the state of the art in every one of the 20 multimodal benchmarks we examined. We believe that the new capabilities of Gemini models in cross-modal reasoning and language understanding will enable a wide variety of use cases and we discuss our approach toward deploying them responsibly to users.

Post-processing mitigation techniques for group fairness generally adjust the decision threshold of a base model in order to improve fairness. Methods in this family exhibit several advantages that make them appealing in practice: post-processing requires no access to the model training pipeline, is agnostic to the base model architecture, and offers a reduced computation cost compared to in-processing. Despite these benefits, existing methods face other challenges that limit their applicability: they require knowledge of the sensitive attributes at inference time and are oftentimes outperformed by in-processing. In this paper, we propose a general framework to transform any in-processing method with a penalized objective into a post-processing procedure. The resulting method is specifically designed to overcome the aforementioned shortcomings of prior post-processing approaches. Furthermore, we show theoretically and through extensive experiments on real-world data that the resulting post-processing method matches or even surpasses the fairness-error trade-off offered by the in-processing counterpart.

Adversarial testing of large language models (LLMs) is crucial for their safe and responsible deployment. We introduce a novel approach for automated generation of adversarial evaluation datasets to test the safety of LLM generations on new downstream applications. We call it AI-assisted Red-Teaming (AART) - an automated alternative to current manual red-teaming efforts. AART offers a data generation and augmentation pipeline of reusable and customizable recipes that reduce human effort significantly and enable integration of adversarial testing earlier in new product development. AART generates evaluation datasets with high diversity of content characteristics critical for effective adversarial testing (e.g. sensitive and harmful concepts, specific to a wide range of cultural and geographic regions and application scenarios). The data generation is steered by AI-assisted recipes to define, scope and prioritize diversity within the application context. This feeds into a structured LLM-generation process that scales up evaluation priorities. Compared to some state-of-the-art tools, AART shows promising results in terms of concept coverage and data quality.

A crucial challenge for generative large language models (LLMs) is diversity: when a user's prompt is under-specified, models may follow implicit assumptions while generating a response, which may result in homogenization of the responses, as well as certain demographic groups being under-represented or even erased from the generated responses. In this paper, we formalize diversity of representation in generative LLMs. We present evaluation datasets and propose metrics to measure diversity in generated responses along people and culture axes. We find that LLMs understand the notion of diversity, and that they can reason and critique their own responses for that goal. This finding motivated a new prompting technique called collective-critique and self-voting (CCSV) to self-improve people diversity of LLMs by tapping into its diversity reasoning capabilities, without relying on handcrafted examples or prompt tuning. Extensive empirical experiments with both human and automated evaluations show that our proposed approach is effective at improving people and culture diversity, and outperforms all baseline methods by a large margin.

Reliably predicting potential failure risks of machine learning (ML) systems when deployed with production data is a crucial aspect of trustworthy AI. This paper introduces Risk Advisor, a novel post-hoc meta-learner for estimating failure risks and predictive uncertainties of any already-trained black-box classification model. In addition to providing a risk score, the Risk Advisor decomposes the uncertainty estimates into aleatoric and epistemic uncertainty components, thus giving informative insights into the sources of uncertainty inducing the failures. Consequently, Risk Advisor can distinguish between failures caused by data variability, data shifts and model limitations and advise on mitigation actions (e.g., collecting more data to counter data shift). Extensive experiments on various families of black-box classification models and on real-world and synthetic datasets covering common ML failure scenarios show that the Risk Advisor reliably predicts deployment-time failure risks in all the scenarios, and outperforms strong baselines.

Much of the previous machine learning (ML) fairness literature assumes that protected features such as race and sex are present in the dataset, and relies upon them to mitigate fairness concerns. However, in practice factors like privacy and regulation often preclude the collection of protected features, or their use for training or inference, severely limiting the applicability of traditional fairness research. Therefore we ask: How can we train an ML model to improve fairness when we do not even know the protected group memberships? In this work we address this problem by proposing Adversarially Reweighted Learning (ARL). In particular, we hypothesize that non-protected features and task labels are valuable for identifying fairness issues, and can be used to co-train an adversarial reweighting approach for improving fairness. Our results show that {ARL} improves Rawlsian Max-Min fairness, with notable AUC improvements for worst-case protected groups in multiple datasets, outperforming state-of-the-art alternatives.

The notion of individual fairness requires that similar people receive similar treatment. However, this is hard to achieve in practice since it is difficult to specify the appropriate similarity metric. In this work, we attempt to learn such similarity metric from human annotated data. We gather a new dataset of human judgments on a criminal recidivism prediction (COMPAS) task. By assuming the human supervision obeys the principle of individual fairness, we leverage prior work on metric learning, evaluate the performance of several metric learning methods on our dataset, and show that the learned metrics outperform the Euclidean and Precision metric under various criteria. We do not provide a way to directly learn a similarity metric satisfying the individual fairness, but to provide an empirical study on how to derive the similarity metric from human supervisors, then future work can use this as a tool to understand human supervision.

We revisit the notion of individual fairness proposed by Dwork et al. A central challenge in operationalizing their approach is the difficulty in eliciting a human specification of a similarity metric. In this paper, we propose an operationalization of individual fairness that does not rely on a human specification of a distance metric. Instead, we propose novel approaches to elicit and leverage side-information on equally deserving individuals to counter subordination between social groups. We model this knowledge as a fairness graph, and learn a unified Pairwise Fair Representation(PFR) of the data that captures both data-driven similarity between individuals and the pairwise side-information in fairness graph. We elicit fairness judgments from a variety of sources, including humans judgments for two real-world datasets on recidivism prediction (COMPAS) and violent neighborhood prediction (Crime & Communities). Our experiments show that the PFR model for operationalizing individual fairness is practically viable.

Abstract--People are rated and ranked, towards algorithmic decision making in an increasing number of applications, typically basedon machine learning. Research on how to incorporate fairness into such tasks has prevalently pursued the paradigm of group fairness: giving adequate success rates to specifically protected groups. In contrast, the alternative paradigm of individual fairnesshas received relatively little attention, and this paper advances this less explored direction. The paper introduces a method for probabilistically mapping user records into a lowrank representationthat reconciles individual fairness and the utility of classifiers and rankings in downstream applications. Our notion of individual fairness requires that users who are similar in all task-relevant attributes such as job qualification, and disregarding all potentially discriminating attributes such as gender, should have similar outcomes. We demonstrate the versatility of our method by applying it to classification and learning-to-rank tasks on a variety of real-world datasets. Our experiments show substantial improvements over the best prior work for this setting. This is a preprint of a full paper at ICDE 2019. Please cite the ICDE proceedings version. I. INTRODUCTION Motivation: People are rated, ranked and selected or not selected inan increasing number of online applications, towards algorithmic decisions based on machine learning models. Examples are approvals or denials of loans or visas, predicting recidivism for law enforcement, or rankings in job portals. As algorithmic decision making becomes pervasive in all aspects of our daily life, societal and ethical concerns [1, 6] are rapidly growing. A basic approach is to establish policies that disallow the inclusion of potentially discriminating attributes such as gender or race, and ensure that classifiers and rankings operate solely on task-relevant attributes such as job qualifications.