arXiv subjects

Angelina Wang

Publications and source records attributed to Angelina Wang.

At least 37 records · Page 2Linked to original sources

arXiv2025

Fairness through Difference Awareness: Measuring Desired Group Discrimination in LLMs

Algorithmic fairness has conventionally adopted the mathematically convenient perspective of racial color-blindness (i.e., difference unaware treatment). However, we contend that in a range of important settings, group difference awareness matters. For example, differentiating between groups may be necessary in legal contexts (e.g., the U.S. compulsory draft applies to men but not women) and harm assessments (e.g., referring to girls as ``terrorists'' may be less harmful than referring to Muslim people as such). Thus, in contrast to most fairness work, we study fairness through the perspective of treating people differently -- when it is contextually appropriate to. We first introduce an important distinction between descriptive (fact-based), normative (value-based), and correlation (association-based) benchmarks. This distinction is significant because each category requires separate interpretation and mitigation tailored to its specific characteristics. Then, we present a benchmark suite composed of eight different scenarios for a total of 16k questions that enables us to assess difference awareness. Finally, we show results across ten models that demonstrate difference awareness is a distinct dimension to fairness where existing bias mitigation strategies may backfire.

cs.CY↗

arXiv2025

Measurement to Meaning: A Validity-Centered Framework for AI Evaluation

While the capabilities and utility of AI systems have advanced, rigorous norms for evaluating these systems have lagged. Grand claims, such as models achieving general reasoning capabilities, are supported with model performance on narrow benchmarks, like performance on graduate-level exam questions, which provide a limited and potentially misleading assessment. We provide a structured approach for reasoning about the types of evaluative claims that can be made given the available evidence. For instance, our framework helps determine whether performance on a mathematical benchmark is an indication of the ability to solve problems on math tests or instead indicates a broader ability to reason. Our framework is well-suited for the contemporary paradigm in machine learning, where various stakeholders provide measurements and evaluations that downstream users use to validate their claims and decisions. At the same time, our framework also informs the construction of evaluations designed to speak to the validity of the relevant claims. By leveraging psychometrics' breakdown of validity, evaluations can prioritize the most critical facets for a given claim, improving empirical utility and decision-making efficacy. We illustrate our framework through detailed case studies of vision and language model evaluations, highlighting how explicitly considering validity strengthens the connection between evaluation evidence and the claims being made.

cs.CY↗

arXiv2025

The California Report on Frontier AI Policy

The innovations emerging at the frontier of artificial intelligence (AI) are poised to create historic opportunities for humanity but also raise complex policy challenges. Continued progress in frontier AI carries the potential for profound advances in scientific discovery, economic productivity, and broader social well-being. As the epicenter of global AI innovation, California has a unique opportunity to continue supporting developments in frontier AI while addressing substantial risks that could have far reaching consequences for the state and beyond. This report leverages broad evidence, including empirical research, historical analysis, and modeling and simulations, to provide a framework for policymaking on the frontier of AI development. Building on this multidisciplinary approach, this report derives policy principles that can inform how California approaches the use, assessment, and governance of frontier AI: principles rooted in an ethos of trust but verify. This approach takes into account the importance of innovation while establishing appropriate strategies to reduce material risks.

cs.CY↗

arXiv2025

Position: Evaluating Generative AI Systems Is a Social Science Measurement Challenge

The measurement tasks involved in evaluating generative AI (GenAI) systems lack sufficient scientific rigor, leading to what has been described as "a tangle of sloppy tests [and] apples-to-oranges comparisons" (Roose, 2024). In this position paper, we argue that the ML community would benefit from learning from and drawing on the social sciences when developing and using measurement instruments for evaluating GenAI systems. Specifically, our position is that evaluating GenAI systems is a social science measurement challenge. We present a four-level framework, grounded in measurement theory from the social sciences, for measuring concepts related to the capabilities, behaviors, and impacts of GenAI systems. This framework has two important implications: First, it can broaden the expertise involved in evaluating GenAI systems by enabling stakeholders with different perspectives to participate in conceptual debates. Second, it brings rigor to both conceptual and operational debates by offering a set of lenses for interrogating validity.

cs.CY↗

arXiv2025

Measuring Machine Learning Harms from Stereotypes Requires Understanding Who Is Harmed by Which Errors in What Ways

As machine learning applications proliferate, we need an understanding of their potential for harm. However, current fairness metrics are rarely grounded in human psychological experiences of harm. Drawing on the social psychology of stereotypes, we use a case study of gender stereotypes in image search to examine how people react to machine learning errors. First, we use survey studies to show that not all machine learning errors reflect stereotypes nor are equally harmful. Then, in experimental studies we randomly expose participants to stereotype-reinforcing, -violating, and -neutral machine learning errors. We find stereotype-reinforcing errors induce more experientially (i.e., subjectively) harmful experiences, while having minimal changes to cognitive beliefs, attitudes, or behaviors. This experiential harm impacts women more than men. However, certain stereotype-violating errors are more experientially harmful for men, potentially due to perceived threats to masculinity. We conclude that harm cannot be the sole guide in fairness mitigation, and propose a nuanced perspective depending on who is experiencing what harm and why.

cs.CY↗

arXiv2025

Toward an Evaluation Science for Generative AI Systems

There is an increasing imperative to anticipate and understand the performance and safety of generative AI systems in real-world deployment contexts. However, the current evaluation ecosystem is insufficient: Commonly used static benchmarks face validity challenges, and ad hoc case-by-case audits rarely scale. In this piece, we advocate for maturing an evaluation science for generative AI systems. While generative AI creates unique challenges for system safety engineering and measurement science, the field can draw valuable insights from the development of safety evaluation practices in other fields, including transportation, aerospace, and pharmaceutical engineering. In particular, we present three key lessons: Evaluation metrics must be applicable to real-world performance, metrics must be iteratively refined, and evaluation institutions and norms must be established. Applying these insights, we outline a concrete path toward a more rigorous approach for evaluating generative AI systems.

cs.AI↗

arXiv2025

Large language models that replace human participants can harmfully misportray and flatten identity groups

Large language models (LLMs) are increasing in capability and popularity, propelling their application in new domains -- including as replacements for human participants in computational social science, user testing, annotation tasks, and more. In many settings, researchers seek to distribute their surveys to a sample of participants that are representative of the underlying human population of interest. This means in order to be a suitable replacement, LLMs will need to be able to capture the influence of positionality (i.e., relevance of social identities like gender and race). However, we show that there are two inherent limitations in the way current LLMs are trained that prevent this. We argue analytically for why LLMs are likely to both misportray and flatten the representations of demographic groups, then empirically show this on 4 LLMs through a series of human studies with 3200 participants across 16 demographic identities. We also discuss a third limitation about how identity prompts can essentialize identities. Throughout, we connect each limitation to a pernicious history of epistemic injustice against the value of lived experiences that explains why replacement is harmful for marginalized demographic groups. Overall, we urge caution in use cases where LLMs are intended to replace human participants whose identities are relevant to the task at hand. At the same time, in cases where the benefits of LLM replacement are determined to outweigh the harms (e.g., the goal is to supplement rather than fully replace, engaging human participants may cause them harm), we provide inference-time techniques that we empirically demonstrate do reduce, but do not remove, these harms.

cs.CY↗

arXiv2024

Evaluating Generative AI Systems is a Social Science Measurement Challenge

Across academia, industry, and government, there is an increasing awareness that the measurement tasks involved in evaluating generative AI (GenAI) systems are especially difficult. We argue that these measurement tasks are highly reminiscent of measurement tasks found throughout the social sciences. With this in mind, we present a framework, grounded in measurement theory from the social sciences, for measuring concepts related to the capabilities, impacts, opportunities, and risks of GenAI systems. The framework distinguishes between four levels: the background concept, the systematized concept, the measurement instrument(s), and the instance-level measurements themselves. This four-level approach differs from the way measurement is typically done in ML, where researchers and practitioners appear to jump straight from background concepts to measurement instruments, with little to no explicit systematization in between. As well as surfacing assumptions, thereby making it easier to understand exactly what the resulting measurements do and do not mean, this framework has two important implications for evaluating evaluations: First, it can enable stakeholders from different worlds to participate in conceptual debates, broadening the expertise involved in evaluating GenAI systems. Second, it brings rigor to operational debates by offering a set of lenses for interrogating the validity of measurement instruments and their resulting measurements.

cs.CY↗

arXiv2024

Strategies for Increasing Corporate Responsible AI Prioritization

Responsible artificial intelligence (RAI) is increasingly recognized as a critical concern. However, the level of corporate RAI prioritization has not kept pace. In this work, we conduct 16 semi-structured interviews with practitioners to investigate what has historically motivated companies to increase the prioritization of RAI. What emerges is a complex story of conflicting and varied factors, but we bring structure to the narrative by highlighting the different strategies available to employ, and point to the actors with access to each. While there are no guaranteed steps for increasing RAI prioritization, we paint the current landscape of motivators so that practitioners can learn from each other, and put forth our own selection of promising directions forward.

cs.CY↗

arXiv2024

Visions of a Discipline: Analyzing Introductory AI Courses on YouTube

Education plays an indispensable role in fostering societal well-being and is widely regarded as one of the most influential factors in shaping the future of generations to come. As artificial intelligence (AI) becomes more deeply integrated into our daily lives and the workforce, educational institutions at all levels are directing their focus on resources that cater to AI education. Our work investigates the current landscape of introductory AI courses on YouTube, and the potential for introducing ethics in this context. We qualitatively analyze the 20 most watched introductory AI courses on YouTube, coding a total of 92.2 hours of educational content viewed by close to 50 million people. Introductory AI courses do not meaningfully engage with ethical or societal challenges of AI (RQ1). When \textit{defining and framing AI}, introductory AI courses foreground excitement around AI's transformative role in society, over-exaggerate AI's current and future abilities, and anthropomorphize AI (RQ2). In \textit{teaching AI}, we see a widespread reliance on corporate AI tools and frameworks as well as a prioritization on a hands-on approach to learning rather than on conceptual foundations (RQ3). In promoting key \textit{AI practices}, introductory AI courses abstract away entirely the socio-technical nature of AI classification and prediction, for example by favoring data quantity over data quality (RQ4). We extend our analysis with recommendations that aim to integrate ethical reflections into introductory AI courses. We recommend that introductory AI courses should (1) highlight ethical challenges of AI to present a more balanced perspective, (2) raise ethical issues explicitly relevant to the technical concepts discussed and (3) nurture a sense of accountability in future AI developers.

cs.CY↗

arXiv2024

Measuring Implicit Bias in Explicitly Unbiased Large Language Models

Large language models (LLMs) can pass explicit social bias tests but still harbor implicit biases, similar to humans who endorse egalitarian beliefs yet exhibit subtle biases. Measuring such implicit biases can be a challenge: as LLMs become increasingly proprietary, it may not be possible to access their embeddings and apply existing bias measures; furthermore, implicit biases are primarily a concern if they affect the actual decisions that these systems make. We address both challenges by introducing two new measures of bias: LLM Implicit Bias, a prompt-based method for revealing implicit bias; and LLM Decision Bias, a strategy to detect subtle discrimination in decision-making tasks. Both measures are based on psychological research: LLM Implicit Bias adapts the Implicit Association Test, widely used to study the automatic associations between concepts held in human minds; and LLM Decision Bias operationalizes psychological results indicating that relative evaluations between two candidates, not absolute evaluations assessing each independently, are more diagnostic of implicit biases. Using these measures, we found pervasive stereotype biases mirroring those in society in 8 value-aligned models across 4 social categories (race, gender, religion, health) in 21 stereotypes (such as race and criminality, race and weapons, gender and science, age and negativity). Our prompt-based LLM Implicit Bias measure correlates with existing language model embedding-based bias methods, but better predicts downstream behaviors measured by LLM Decision Bias. These new prompt-based measures draw from psychology's long history of research into measuring stereotype biases based on purely observable behavior; they expose nuanced biases in proprietary value-aligned LLMs that appear unbiased according to standard benchmarks.

cs.CY↗

arXiv2023

Gender Artifacts in Visual Datasets

Gender biases are known to exist within large-scale visual datasets and can be reflected or even amplified in downstream models. Many prior works have proposed methods for mitigating gender biases, often by attempting to remove gender expression information from images. To understand the feasibility and practicality of these approaches, we investigate what $\textit{gender artifacts}$ exist within large-scale visual datasets. We define a $\textit{gender artifact}$ as a visual cue that is correlated with gender, focusing specifically on those cues that are learnable by a modern image classifier and have an interpretable human corollary. Through our analyses, we find that gender artifacts are ubiquitous in the COCO and OpenImages datasets, occurring everywhere from low-level information (e.g., the mean value of the color channels) to the higher-level composition of the image (e.g., pose and location of people). Given the prevalence of gender artifacts, we claim that attempts to remove gender artifacts from such datasets are largely infeasible. Instead, the responsibility lies with researchers and practitioners to be aware that the distribution of images within datasets is highly gendered and hence develop methods which are robust to these distributional shifts across groups.

cs.CV↗

arXiv2023

Overwriting Pretrained Bias with Finetuning Data

Transfer learning is beneficial by allowing the expressive features of models pretrained on large-scale datasets to be finetuned for the target task of smaller, more domain-specific datasets. However, there is a concern that these pretrained models may come with their own biases which would propagate into the finetuned model. In this work, we investigate bias when conceptualized as both spurious correlations between the target task and a sensitive attribute as well as underrepresentation of a particular group in the dataset. Under both notions of bias, we find that (1) models finetuned on top of pretrained models can indeed inherit their biases, but (2) this bias can be corrected for through relatively minor interventions to the finetuning dataset, and often with a negligible impact to performance. Our findings imply that careful curation of the finetuning dataset is important for reducing biases on a downstream task, and doing so can even compensate for bias in the pretrained model.

cs.CV↗

arXiv2023

Taxonomizing and Measuring Representational Harms: A Look at Image Tagging

In this paper, we examine computational approaches for measuring the "fairness" of image tagging systems, finding that they cluster into five distinct categories, each with its own analytic foundation. We also identify a range of normative concerns that are often collapsed under the terms "unfairness," "bias," or even "discrimination" when discussing problematic cases of image tagging. Specifically, we identify four types of representational harms that can be caused by image tagging systems, providing concrete examples of each. We then consider how different computational measurement approaches map to each of these types, demonstrating that there is not a one-to-one mapping. Our findings emphasize that no single measurement approach will be definitive and that it is not possible to infer from the use of a particular measurement approach which type of harm was intended to be measured. Lastly, equipped with this more granular understanding of the types of representational harms that can be caused by image tagging systems, we show that attempts to mitigate some of these types of harms may be in tension with one another.

cs.CY↗

arXiv2022

Measuring Representational Harms in Image Captioning

Previous work has largely considered the fairness of image captioning systems through the underspecified lens of "bias." In contrast, we present a set of techniques for measuring five types of representational harms, as well as the resulting measurements obtained for two of the most popular image captioning datasets using a state-of-the-art image captioning system. Our goal was not to audit this image captioning system, but rather to develop normatively grounded measurement techniques, in turn providing an opportunity to reflect on the many challenges involved. We propose multiple measurement techniques for each type of harm. We argue that by doing so, we are better able to capture the multi-faceted nature of each type of harm, in turn improving the (collective) validity of the resulting measurements. Throughout, we discuss the assumptions underlying our measurement approach and point out when they do not hold.

cs.CY↗

arXiv2022

Towards Intersectionality in Machine Learning: Including More Identities, Handling Underrepresentation, and Performing Evaluation

Research in machine learning fairness has historically considered a single binary demographic attribute; however, the reality is of course far more complicated. In this work, we grapple with questions that arise along three stages of the machine learning pipeline when incorporating intersectionality as multiple demographic attributes: (1) which demographic attributes to include as dataset labels, (2) how to handle the progressively smaller size of subgroups during model training, and (3) how to move beyond existing evaluation metrics when benchmarking model fairness for more subgroups. For each question, we provide thorough empirical evaluation on tabular datasets derived from the US Census, and present constructive recommendations for the machine learning community. First, we advocate for supplementing domain knowledge with empirical validation when choosing which demographic attribute labels to train on, while always evaluating on the full set of demographic attributes. Second, we warn against using data imbalance techniques without considering their normative implications and suggest an alternative using the structure in the data. Third, we introduce new evaluation metrics which are more appropriate for the intersectional setting. Overall, we provide substantive suggestions on three necessary (albeit not sufficient!) considerations when incorporating intersectionality into machine learning.

cs.LG↗

arXiv2021

Understanding and Evaluating Racial Biases in Image Captioning

Image captioning is an important task for benchmarking visual reasoning and for enabling accessibility for people with vision impairments. However, as in many machine learning settings, social biases can influence image captioning in undesirable ways. In this work, we study bias propagation pathways within image captioning, focusing specifically on the COCO dataset. Prior work has analyzed gender bias in captions using automatically-derived gender labels; here we examine racial and intersectional biases using manual annotations. Our first contribution is in annotating the perceived gender and skin color of 28,315 of the depicted people after obtaining IRB approval. Using these annotations, we compare racial biases present in both manual and automatically-generated image captions. We demonstrate differences in caption performance, sentiment, and word choice between images of lighter versus darker-skinned people. Further, we find the magnitude of these differences to be greater in modern captioning systems compared to older ones, thus leading to concerns that without proper consideration and mitigation these differences will only become increasingly prevalent. Code and data is available at https://princetonvisualai.github.io/imagecaptioning-bias .

cs.CV↗

arXiv2021

REVISE: A Tool for Measuring and Mitigating Bias in Visual Datasets

Machine learning models are known to perpetuate and even amplify the biases present in the data. However, these data biases frequently do not become apparent until after the models are deployed. Our work tackles this issue and enables the preemptive analysis of large-scale datasets. REVISE (REvealing VIsual biaSEs) is a tool that assists in the investigation of a visual dataset, surfacing potential biases along three dimensions: (1) object-based, (2) person-based, and (3) geography-based. Object-based biases relate to the size, context, or diversity of the depicted objects. Person-based metrics focus on analyzing the portrayal of people within the dataset. Geography-based analyses consider the representation of different geographic locations. These three dimensions are deeply intertwined in how they interact to bias a dataset, and REVISE sheds light on this; the responsibility then lies with the user to consider the cultural and historical context, and to determine which of the revealed biases may be problematic. The tool further assists the user by suggesting actionable steps that may be taken to mitigate the revealed biases. Overall, the key aim of our work is to tackle the machine learning bias problem early in the pipeline. REVISE is available at https://github.com/princetonvisualai/revise-tool

cs.CV↗