Kenneth Holstein’s research while affiliated with Carnegie Mellon University and other places

The LLM-as-a-judge paradigm, in which a judge LLM system replaces human raters in rating the outputs of other generative AI (GenAI) systems, has come to play a critical role in scaling and standardizing GenAI evaluations. To validate judge systems, evaluators collect multiple human ratings for each item in a validation corpus, and then aggregate the ratings into a single, per-item gold label rating. High agreement rates between these gold labels and judge system ratings are then taken as a sign of good judge system performance. In many cases, however, items or rating criteria may be ambiguous, or there may be principled disagreement among human raters. In such settings, gold labels may not exist for many of the items. In this paper, we introduce a framework for LLM-as-a-judge validation in the absence of gold labels. We present a theoretical analysis drawing connections between different measures of judge system performance under different rating elicitation and aggregation schemes. We also demonstrate empirically that existing validation approaches can select judge systems that are highly suboptimal, performing as much as 34% worse than the systems selected by alternative approaches that we describe. Based on our findings, we provide concrete recommendations for developing more reliable approaches to LLM-as-a-judge validation.

AI systems are often introduced with high expectations, yet many fail to deliver, resulting in unintended harm and missed opportunities for benefit. We frequently observe significant "AI Mismatches", where the system's actual performance falls short of what is needed to ensure safety and co-create value. These mismatches are particularly difficult to address once development is underway, highlighting the need for early-stage intervention. Navigating complex, multi-dimensional risk factors that contribute to AI Mismatches is a persistent challenge. To address it, we propose an AI Mismatch approach to anticipate and mitigate risks early on, focusing on the gap between realistic model performance and required task performance. Through an analysis of 774 AI cases, we extracted a set of critical factors, which informed the development of seven matrices that map the relationships between these factors and highlight high-risk areas. Through case studies, we demonstrate how our approach can help reduce risks in AI development.

Despite Generative AI (GenAI) systems' potential for enhancing content creation, users often struggle to effectively integrate GenAI into their creative workflows. Core challenges include misalignment of AI-generated content with user intentions (intent elicitation and alignment), user uncertainty around how to best communicate their intents to the AI system (prompt formulation), and insufficient flexibility of AI systems to support diverse creative workflows (workflow flexibility). Motivated by these challenges, we created IntentTagger: a system for slide creation based on the notion of Intent Tags - small, atomic conceptual units that encapsulate user intent - for exploring granular and non-linear micro-prompting interactions for Human-GenAI co-creation workflows. Our user study with 12 participants provides insights into the value of flexibly expressing intent across varying levels of ambiguity, meta-intent elicitation, and the benefits and challenges of intent tag-driven workflows. We conclude by discussing the broader implications of our findings and design considerations for GenAI-supported content creation workflows.

There has been growing interest from both practitioners and researchers in engaging end users in AI auditing, to draw upon users’ unique knowledge and lived experiences. However, we know little about how to effectively scaffold end users in auditing in ways that can generate actionable insights for AI practitioners. Through formative studies with both users and AI practitioners, we first identified a set of design goals to support user-engaged AI auditing. We then developed _WeAudit_, a workflow and system that supports end users in auditing AI both individually and collectively. We evaluated _WeAudit_ through a three-week user study with user auditors and interviews with industry Generative AI practitioners. Our findings offer insights into how _WeAudit_ supports users in noticing and reflecting upon potential AI harms and in articulating their findings in ways that industry practitioners can act upon. Based on our observations and feedback from both users and practitioners, we identify several opportunities to better support user engagement in AI auditing processes. We discuss implications for future research to support effective and responsible user engagement in AI auditing.

There has been growing interest from both practitioners and researchers in engaging end users in AI auditing, to draw upon users' unique knowledge and lived experiences. However, we know little about how to effectively scaffold end users in auditing in ways that can generate actionable insights for AI practitioners. Through formative studies with both users and AI practitioners, we first identified a set of design goals to support user-engaged AI auditing. We then developed WeAudit, a workflow and system that supports end users in auditing AI both individually and collectively. We evaluated WeAudit through a three-week user study with user auditors and interviews with industry Generative AI practitioners. Our findings offer insights into how WeAudit supports users in noticing and reflecting upon potential AI harms and in articulating their findings in ways that industry practitioners can act upon. Based on our observations and feedback from both users and practitioners, we identify several opportunities to better support user engagement in AI auditing processes. We discuss implications for future research to support effective and responsible user engagement in AI auditing and red-teaming.

As public sector agencies rapidly introduce new AI tools in high-stakes domains like social services, it becomes critical to understand how decisions to adopt these tools are made in practice. We borrow from the anthropological practice to "study up" those in positions of power, and reorient our study of public sector AI around those who have the power and responsibility to make decisions about the role that AI tools will play in their agency. Through semi-structured interviews and design activities with 16 agency decision-makers, we examine how decisions about AI design and adoption are influenced by their interactions with and assumptions about other actors within these agencies (e.g., frontline workers and agency leaders), as well as those above (legal systems and contracted companies), and below (impacted communities). By centering these networks of power relations, our findings shed light on how infrastructural, legal, and social factors create barriers and disincentives to the involvement of a broader range of stakeholders in decisions about AI design and adoption. Agency decision-makers desired more practical support for stakeholder involvement around public sector AI to help overcome the knowledge and power differentials they perceived between them and other stakeholders (e.g., frontline workers and impacted community members). Building on these findings, we discuss implications for future research and policy around actualizing participatory AI approaches in public sector contexts.

With the rise of generative AI (GenAI), there has been an increased need for participation by large and diverse user bases in AI evaluation and auditing. GenAI developers are increasingly adopting crowdsourcing approaches to test and audit their AI products and services. However, it remains an open question how to design and deploy responsible and effective crowdsourcing pipelines for AI auditing and evaluation. This workshop aims to take a step towards bridging this gap. Our interdisciplinary team of organizers will work with workshop participants to explore several key questions, such as how to improve the output quality and workers' productivity for GenAI evaluation crowdsourcing tasks compared to discriminative AI systems, how to guide crowds in auditing problematic AI-generated content while managing their psychological impact, ensuring marginalized voices are heard, and setting up responsible and effective crowdsourcing pipelines for real-world GenAI evaluation. We hope this workshop will produce a research agenda and best practices for designing responsible crowd-based approaches to AI auditing and evaluation.

There has been growing recognition of the crucial role users, especially those from marginalized groups, play in uncovering harmful algorithmic biases. However, it remains unclear how users’ identities and experiences might impact their rating of harmful biases. We present an online experiment (N=2,197) examining these factors: demographics, discrimination experiences, and social and technical knowledge. Participants were shown examples of image search results, including ones that previous literature has identified as biased against marginalized racial, gender, or sexual orientation groups. We found participants from marginalized gender or sexual orientation groups were more likely to rate the examples as more severely harmful. Belonging to marginalized races did not have a similar pattern. Additional factors affecting users’ ratings included discrimination experiences, and having friends or family belonging to marginalized demographics. A qualitative analysis offers insights into users' bias recognition, and why they see biases the way they do. We provide guidance for designing future methods to support effective user-driven auditing.