Kenneth Holstein

Wesley Hanwen Deng, Manish Nagireddy, Michelle Seng Ah Lee et al. · cmu

Recent years have seen the development of many open-source ML fairness toolkits aimed at helping ML practitioners assess and address unfairness in their systems. However, there has been little research investigating how ML practitioners actually use these toolkits in practice. In this paper, we conducted the first in-depth empirical exploration of how industry practitioners (try to) work with existing fairness toolkits. In particular, we conducted think-aloud interviews to understand how participants learn about and use fairness toolkits, and explored the generality of our findings through an anonymous online survey. We identified several opportunities for fairness toolkits to better address practitioner needs and scaffold them in using toolkits effectively and responsibly. Based on these findings, we highlight implications for the design of future open-source fairness toolkits that can support practitioners in better contextualizing, communicating, and collaborating around ML fairness efforts.

Frederic Gmeiner, Humphrey Yang, Lining Yao et al. · cmu

AI-based design tools are proliferating in professional software to assist engineering and industrial designers in complex manufacturing and design tasks. These tools take on more agentic roles than traditional computer-aided design tools and are often portrayed as "co-creators." Yet, working effectively with such systems requires different skills than working with complex CAD tools alone. To date, we know little about how engineering designers learn to work with AI-based design tools. In this study, we observed trained designers as they learned to work with two AI-based tools on a realistic design task. We find that designers face many challenges in learning to effectively co-create with current systems, including challenges in understanding and adjusting AI outputs and in communicating their design goals. Based on our findings, we highlight several design opportunities to better support designer-AI co-creation.

Anna Kawakami, Venkatesh Sivaraman, Hao-Fei Cheng et al. · cmu

AI-based decision support tools (ADS) are increasingly used to augment human decision-making in high-stakes, social contexts. As public sector agencies begin to adopt ADS, it is critical that we understand workers' experiences with these systems in practice. In this paper, we present findings from a series of interviews and contextual inquiries at a child welfare agency, to understand how they currently make AI-assisted child maltreatment screening decisions. Overall, we observe how workers' reliance upon the ADS is guided by (1) their knowledge of rich, contextual information beyond what the AI model captures, (2) their beliefs about the ADS's capabilities and limitations relative to their own, (3) organizational pressures and incentives around the use of the ADS, and (4) awareness of misalignments between algorithmic predictions and their own decision-making objectives. Drawing upon these findings, we discuss design implications towards supporting more effective human-AI decision-making.

Wesley Hanwen Deng, Bill Boyuan Guo, Alicia DeVrio et al. · cmu

Recent years have seen growing interest among both researchers and practitioners in user-engaged approaches to algorithm auditing, which directly engage users in detecting problematic behaviors in algorithmic systems. However, we know little about industry practitioners' current practices and challenges around user-engaged auditing, nor what opportunities exist for them to better leverage such approaches in practice. To investigate, we conducted a series of interviews and iterative co-design activities with practitioners who employ user-engaged auditing approaches in their work. Our findings reveal several challenges practitioners face in appropriately recruiting and incentivizing user auditors, scaffolding user audits, and deriving actionable insights from user-engaged audit reports. Furthermore, practitioners shared organizational obstacles to user-engaged auditing, surfacing a complex relationship between practitioners and user auditors. Based on these findings, we discuss opportunities for future HCI research to help realize the potential (and the mitigate risks) of user-engaged auditing in industry practice.

Anna Kawakami, Luke Guerdan, Yanghuidi Cheng et al. · cmu

A growing body of research has explored how to support humans in making better use of AI-based decision support, including via training and onboarding. Existing research has focused on decision-making tasks where it is possible to evaluate "appropriate reliance" by comparing each decision against a ground truth label that cleanly maps to both the AI's predictive target and the human decision-maker's goals. However, this assumption does not hold in many real-world settings where AI tools are deployed today (e.g., social work, criminal justice, and healthcare). In this paper, we introduce a process-oriented notion of appropriate reliance called critical use that centers the human's ability to situate AI predictions against knowledge that is uniquely available to them but unavailable to the AI model. To explore how training can support critical use, we conduct a randomized online experiment in a complex social decision-making setting: child maltreatment screening. We find that, by providing participants with accelerated, low-stakes opportunities to practice AI-assisted decision-making in this setting, novices came to exhibit patterns of disagreement with AI that resemble those of experienced workers. A qualitative examination of participants' explanations for their AI-assisted decisions revealed that they drew upon qualitative case narratives, to which the AI model did not have access, to learn when (not) to rely on AI predictions. Our findings open new questions for the study and design of training for real-world AI-assisted decision-making.

Michelle S. Lam, Fred Hohman, Dominik Moritz et al. · apple-ml, cmu

AI policy sets boundaries on acceptable behavior for AI models, but this is challenging in the context of large language models (LLMs): how do you ensure coverage over a vast behavior space? We introduce policy maps, an approach to AI policy design inspired by the practice of physical mapmaking. Instead of aiming for full coverage, policy maps aid effective navigation through intentional design choices about which aspects to capture and which to abstract away. With Policy Projector, an interactive tool for designing LLM policy maps, an AI practitioner can survey the landscape of model input-output pairs, define custom regions (e.g., "violence"), and navigate these regions with if-then policy rules that can act on LLM outputs (e.g., if output contains "violence" and "graphic details," then rewrite without "graphic details"). Policy Projector supports interactive policy authoring using LLM classification and steering and a map visualization reflecting the AI practitioner's work. In an evaluation with 12 AI safety experts, our system helps policy designers craft policies around problematic model behaviors such as incorrect gender assumptions and handling of immediate physical safety threats.

Frederic Gmeiner, Kenneth Holstein, Nikolas Martelaro · cmu

Generative, ML-driven interactive systems have the potential to change how people interact with computers in creative processes - turning tools into co-creators. However, it is still unclear how we might achieve effective human-AI collaboration in open-ended task domains. There are several known challenges around communication in the interaction with ML-driven systems. An overlooked aspect in the design of co-creative systems is how users can be better supported in learning to collaborate with such systems. Here we reframe human-AI collaboration as a learning problem: Inspired by research on team learning, we hypothesize that similar learning strategies that apply to human-human teams might also increase the collaboration effectiveness and quality of humans working with co-creative generative systems. In this position paper, we aim to promote team learning as a lens for designing more effective co-creative human-AI collaboration and emphasize collaboration process quality as a goal for co-creative systems. Furthermore, we outline a preliminary schematic framework for embedding team learning support in co-creative AI systems. We conclude by proposing a research agenda and posing open questions for further study on supporting people in learning to collaborate with generative AI systems.

Anna Kawakami, Amanda Coston, Haiyi Zhu et al. · cmu

AI-based decision-making tools are rapidly spreading across a range of real-world, complex domains like healthcare, criminal justice, and child welfare. A growing body of research has called for increased scrutiny around the validity of AI system designs. However, in real-world settings, it is often not possible to fully address questions around the validity of an AI tool without also considering the design of associated organizational and public policies. Yet, considerations around how an AI tool may interface with policy are often only discussed retrospectively, after the tool is designed or deployed. In this short position paper, we discuss opportunities to promote multi-stakeholder deliberations around the design of AI-based technologies and associated policies, at the earliest stages of a new project.

Rachel M. Kim, Blaine Kuehnert, Alice Lai et al.

As Artificial Intelligence (AI) systems proliferate, the need for systematic, transparent, and actionable processes for evaluating them is growing. While many resources exist to support AI evaluation, they have several limitations. Few address both the process of designing, developing, and deploying an AI system and the outcomes it produces. Furthermore, few are end-to-end and operational, give actionable guidance, or present evidence of usability or effectiveness in practice. In this paper, we introduce a third-party AI assurance framework that addresses these gaps. We focus on third-party assurance to prevent conflict of interest and ensure credibility and accountability of the process. We begin by distinguishing assurance from audits in several key dimensions. Then, following design principles, we reflect on the shortcomings of existing resources to identify a set of design requirements for AI assurance. We then construct a prototype of an assurance process that consists of (1) a responsibility assignment matrix to determine the different levels of involvement each stakeholder has at each stage of the AI lifecycle, (2) an interview protocol for each stakeholder of an AI system, (3) a maturity matrix to assess AI systems' adherence to best practices, and (4) a template for an assurance report that draws from more mature assurance practices in business accounting. We conduct early validation of our AI assurance framework by applying the framework to two distinct AI use cases -- a business document tagging tool for downstream processing in a large private firm, and a housing resource allocation tool in a public agency -- and conducting six expert validation interviews. Our findings show early evidence that our AI assurance framework is sound and comprehensive, usable across different organizational contexts, and effective at identifying bespoke issues with AI systems.

Logan Stapleton, Min Hun Lee, Diana Qing et al.

Child welfare agencies across the United States are turning to data-driven predictive technologies (commonly called predictive analytics) which use government administrative data to assist workers' decision-making. While some prior work has explored impacted stakeholders' concerns with current uses of data-driven predictive risk models (PRMs), less work has asked stakeholders whether such tools ought to be used in the first place. In this work, we conducted a set of seven design workshops with 35 stakeholders who have been impacted by the child welfare system or who work in it to understand their beliefs and concerns around PRMs, and to engage them in imagining new uses of data and technologies in the child welfare system. We found that participants worried current PRMs perpetuate or exacerbate existing problems in child welfare. Participants suggested new ways to use data and data-driven tools to better support impacted communities and suggested paths to mitigate possible harms of these tools. Participants also suggested low-tech or no-tech alternatives to PRMs to address problems in child welfare. Our study sheds light on how researchers and designers can work in solidarity with impacted communities, possibly to circumvent or oppose child welfare agencies.

Luke Guerdan, Amanda Coston, Zhiwei Steven Wu et al.

A growing literature on human-AI decision-making investigates strategies for combining human judgment with statistical models to improve decision-making. Research in this area often evaluates proposed improvements to models, interfaces, or workflows by demonstrating improved predictive performance on "ground truth" labels. However, this practice overlooks a key difference between human judgments and model predictions. Whereas humans reason about broader phenomena of interest in a decision -- including latent constructs that are not directly observable, such as disease status, the "toxicity" of online comments, or future "job performance" -- predictive models target proxy labels that are readily available in existing datasets. Predictive models' reliance on simplistic proxies makes them vulnerable to various sources of statistical bias. In this paper, we identify five sources of target variable bias that can impact the validity of proxy labels in human-AI decision-making tasks. We develop a causal framework to disentangle the relationship between each bias and clarify which are of concern in specific human-AI decision-making tasks. We demonstrate how our framework can be used to articulate implicit assumptions made in prior modeling work, and we recommend evaluation strategies for verifying whether these assumptions hold in practice. We then leverage our framework to re-examine the designs of prior human subjects experiments that investigate human-AI decision-making, finding that only a small fraction of studies examine factors related to target variable bias. We conclude by discussing opportunities to better address target variable bias in future research.

Tzu-Sheng Kuo, Hong Shen, Jisoo Geum et al.

Recent years have seen growing adoption of AI-based decision-support systems (ADS) in homeless services, yet we know little about stakeholder desires and concerns surrounding their use. In this work, we aim to understand impacted stakeholders' perspectives on a deployed ADS that prioritizes scarce housing resources. We employed AI lifecycle comicboarding, an adapted version of the comicboarding method, to elicit stakeholder feedback and design ideas across various components of an AI system's design. We elicited feedback from county workers who operate the ADS daily, service providers whose work is directly impacted by the ADS, and unhoused individuals in the region. Our participants shared concerns and design suggestions around the AI system's overall objective, specific model design choices, dataset selection, and use in deployment. Our findings demonstrate that stakeholders, even without AI knowledge, can provide specific and critical feedback on an AI system's design and deployment, if empowered to do so.

Luke Guerdan, Amanda Coston, Kenneth Holstein et al.

Across domains such as medicine, employment, and criminal justice, predictive models often target labels that imperfectly reflect the outcomes of interest to experts and policymakers. For example, clinical risk assessments deployed to inform physician decision-making often predict measures of healthcare utilization (e.g., costs, hospitalization) as a proxy for patient medical need. These proxies can be subject to outcome measurement error when they systematically differ from the target outcome they are intended to measure. However, prior modeling efforts to characterize and mitigate outcome measurement error overlook the fact that the decision being informed by a model often serves as a risk-mitigating intervention that impacts the target outcome of interest and its recorded proxy. Thus, in these settings, addressing measurement error requires counterfactual modeling of treatment effects on outcomes. In this work, we study intersectional threats to model reliability introduced by outcome measurement error, treatment effects, and selection bias from historical decision-making policies. We develop an unbiased risk minimization method which, given knowledge of proxy measurement error properties, corrects for the combined effects of these challenges. We also develop a method for estimating treatment-dependent measurement error parameters when these are unknown in advance. We demonstrate the utility of our approach theoretically and via experiments on real-world data from randomized controlled trials conducted in healthcare and employment domains. As importantly, we demonstrate that models correcting for outcome measurement error or treatment effects alone suffer from considerable reliability limitations. Our work underscores the importance of considering intersectional threats to model validity during the design and evaluation of predictive models for decision support.

Kenneth Holstein, Maria De-Arteaga, Lakshmi Tumati et al.

In many real world contexts, successful human-AI collaboration requires humans to productively integrate complementary sources of information into AI-informed decisions. However, in practice human decision-makers often lack understanding of what information an AI model has access to in relation to themselves. There are few available guidelines regarding how to effectively communicate about unobservables: features that may influence the outcome, but which are unavailable to the model. In this work, we conducted an online experiment to understand whether and how explicitly communicating potentially relevant unobservables influences how people integrate model outputs and unobservables when making predictions. Our findings indicate that presenting prompts about unobservables can change how humans integrate model outputs and unobservables, but do not necessarily lead to improved performance. Furthermore, the impacts of these prompts can vary depending on decision-makers' prior domain expertise. We conclude by discussing implications for future research and design of AI-based decision support tools.

Philipp Spitzer, Joshua Holstein, Katelyn Morrison et al.

Across various applications, humans increasingly use black-box artificial intelligence (AI) systems without insight into these systems' reasoning. To counter this opacity, explainable AI (XAI) methods promise enhanced transparency and interpretability. While recent studies have explored how XAI affects human-AI collaboration, few have examined the potential pitfalls caused by incorrect explanations. The implications for humans can be far-reaching but have not been explored extensively. To investigate this, we ran a study (n=160) on AI-assisted decision-making in which humans were supported by XAI. Our findings reveal a misinformation effect when incorrect explanations accompany correct AI advice with implications post-collaboration. This effect causes humans to infer flawed reasoning strategies, hindering task execution and demonstrating impaired procedural knowledge. Additionally, incorrect explanations compromise human-AI team-performance during collaboration. With our work, we contribute to HCI by providing empirical evidence for the negative consequences of incorrect explanations on humans post-collaboration and outlining guidelines for designers of AI.

Charvi Rastogi, Liu Leqi, Kenneth Holstein et al.

Hybrid human-ML systems increasingly make consequential decisions in a wide range of domains. These systems are often introduced with the expectation that the combined human-ML system will achieve complementary performance, that is, the combined decision-making system will be an improvement compared with either decision-making agent in isolation. However, empirical results have been mixed, and existing research rarely articulates the sources and mechanisms by which complementary performance is expected to arise. Our goal in this work is to provide conceptual tools to advance the way researchers reason and communicate about human-ML complementarity. Drawing upon prior literature in human psychology, machine learning, and human-computer interaction, we propose a taxonomy characterizing distinct ways in which human and ML-based decision-making can differ. In doing so, we conceptually map potential mechanisms by which combining human and ML decision-making may yield complementary performance, developing a language for the research community to reason about design of hybrid systems in any decision-making domain. To illustrate how our taxonomy can be used to investigate complementarity, we provide a mathematical aggregation framework to examine enabling conditions for complementarity. Through synthetic simulations, we demonstrate how this framework can be used to explore specific aspects of our taxonomy and shed light on the optimal mechanisms for combining human-ML judgments

Wesley Hanwen Deng, Mingxi Yan, Sunnie S. Y. Kim et al.

Recent developments in AI safety research have called for red-teaming methods that effectively surface potential risks posed by generative AI models, with growing emphasis on how red-teamers' backgrounds and perspectives shape their strategies and the risks they uncover. While automated red-teaming approaches promise to complement human red-teaming through larger-scale exploration, existing automated approaches do not account for human identities and rarely incorporate human inputs. In this work, we explore persona-driven red-teaming to advance both automated red-teaming and human-AI collaboration. We first develop PersonaTeaming Workflow, which incorporates personas into the adversarial prompt generation process to explore a wider spectrum of adversarial strategies. Compared to RainbowPlus, a state-of-the-art automated red-teaming method, PersonaTeaming Workflow achieves higher attack success rates while maintaining prompt diversity. However, since automated personas only approximate real human perspectives, we further instantiate PersonaTeaming Workflow as PersonaTeaming Playground, a user-facing interface that enables red-teamers to author their own personas and collaborate with AI to mutate and refine prompts. In a user study with 11 industry practitioners, we found that PersonaTeaming Playground enabled diverse red-teaming strategies and outputs that practitioners perceived as useful, and that AI-generated suggestions in the PersonaTeaming Playground encouraged out-of-the-box thinking even when practitioners did not follow them strictly. Together, our work advances both automated and human-in-the-loop approaches to red-teaming, while shedding light on interaction patterns and design insights for supporting human-AI collaboration in generative AI red-teaming.

Ziyang Guo, Yifan Wu, Jason Hartline et al.

Multi-agent decision pipelines can outperform single agent workflows when complementarity holds, i.e., different agents bring unique information to the table to inform a final decision. We propose ComplLLM, a post-training framework based on decision theory that fine-tunes a decision-assistant LLM using complementary information as reward to output signals that complement existing agent decisions. We validate ComplLLM on synthetic and real-world tasks involving domain experts, demonstrating how the approach recovers known complementary information and produces plausible explanations of complementary signals to support downstream decision-makers.

Tzu-Sheng Kuo, Aaron Halfaker, Zirui Cheng et al.

AI tools are increasingly deployed in community contexts. However, datasets used to evaluate AI are typically created by developers and annotators outside a given community, which can yield misleading conclusions about AI performance. How might we empower communities to drive the intentional design and curation of evaluation datasets for AI that impacts them? We investigate this question on Wikipedia, an online community with multiple AI-based content moderation tools deployed. We introduce Wikibench, a system that enables communities to collaboratively curate AI evaluation datasets, while navigating ambiguities and differences in perspective through discussion. A field study on Wikipedia shows that datasets curated using Wikibench can effectively capture community consensus, disagreement, and uncertainty. Furthermore, study participants used Wikibench to shape the overall data curation process, including refining label definitions, determining data inclusion criteria, and authoring data statements. Based on our findings, we propose future directions for systems that support community-driven data curation.

Anna Kawakami, Amanda Coston, Hoda Heidari et al. · cmu

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.

Frederic Gmeiner, Nicolai Marquardt, Michael Bentley et al.

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.

Devansh Saxena, Ji-Youn Jung, Jodi Forlizzi et al.

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.

Luke Guerdan, Solon Barocas, Kenneth Holstein et al.

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, plays a critical role in scaling and standardizing GenAI evaluations. To validate such judge systems, evaluators assess human--judge agreement by first collecting multiple human ratings for each item in a validation corpus, then aggregating the ratings into a single, per-item gold label rating. For many items, however, rating criteria may admit multiple valid interpretations, so a human or LLM rater may deem multiple ratings "reasonable" or "correct." We call this condition rating indeterminacy. Problematically, many rating tasks that contain rating indeterminacy rely on forced-choice elicitation, whereby raters are instructed to select only one rating for each item. In this paper, we introduce a framework for validating LLM-as-a-judge systems under rating indeterminacy. We draw theoretical connections between different measures of judge system performance under different human--judge agreement metrics, and different rating elicitation and aggregation schemes. We demonstrate that differences in how humans and LLMs resolve rating indeterminacy when responding to forced-choice rating instructions can heavily bias LLM-as-a-judge validation. Through extensive experiments involving 11 real-world rating tasks and 9 commercial LLMs, we show that standard validation approaches that rely upon forced-choice ratings select judge systems that are highly suboptimal, performing as much as 31% worse than judge systems selected by our approach that uses multi-label "response set" ratings to account for rating indeterminacy. We conclude with concrete recommendations for more principled approaches to LLM-as-a-judge validation.

Luke Guerdan, Devansh Saxena, Stevie Chancellor et al.

Data scientists often formulate predictive modeling tasks involving fuzzy, hard-to-define concepts, such as the "authenticity" of student writing or the "healthcare need" of a patient. Yet the process by which data scientists translate fuzzy concepts into a concrete, proxy target variable remains poorly understood. We interview fifteen data scientists in education (N=8) and healthcare (N=7) to understand how they construct target variables for predictive modeling tasks. Our findings suggest that data scientists construct target variables through a bricolage process, in which they use creative and pragmatic approaches to make do with the limited data at hand. Data scientists attempt to satisfy five major criteria for a target variable through bricolage: validity, simplicity, predictability, portability, and resource requirements. To achieve this, data scientists adaptively apply problem (re)formulation strategies, such as swapping out one candidate target variable for another when the first fails to meet certain criteria (e.g., predictability), or composing multiple outcomes into a single target variable to capture a more holistic set of modeling objectives. Based on our findings, we present opportunities for future HCI, CSCW, and ML research to better support the art and science of target variable construction.

Frederic Gmeiner, Kaitao Luo, Ye Wang et al.

Despite the potential of generative AI (GenAI) design tools to enhance design processes, professionals often struggle to integrate AI into their workflows. Fundamental cognitive challenges include the need to specify all design criteria as distinct parameters upfront (intent formulation) and designers' reduced cognitive involvement in the design process due to cognitive offloading, which can lead to insufficient problem exploration, underspecification, and limited ability to evaluate outcomes. Motivated by these challenges, we envision novel metacognitive support agents that assist designers in working more reflectively with GenAI. To explore this vision, we conducted exploratory prototyping through a Wizard of Oz elicitation study with 20 mechanical designers probing multiple metacognitive support strategies. We found that agent-supported users created more feasible designs than non-supported users, with differing impacts between support strategies. Based on these findings, we discuss opportunities and tradeoffs of metacognitive support agents and considerations for future AI-based design tools.

Luke Guerdan, Justin Whitehouse, Kimberly Truong et al.

As Generative AI (GenAI) systems see growing adoption, a key concern involves the external validity of evaluations, or the extent to which they generalize from lab-based to real-world deployment conditions. Threats to the external validity of GenAI evaluations arise when the source sample of human raters and system outputs used to obtain a system quality estimate differs from the target distribution at deployment time. In this work, we propose a doubly-robust estimation framework designed to address this evaluation sampling bias. Key to our approach is the use of "persona" ratings produced by prompting an LLM evaluator (i.e., an LLM-as-a-judge) to behave as a human rater with specific sociodemographic characteristics. Our doubly-robust framework combines these informative yet imperfect persona ratings with human ratings obtained under evaluation sampling bias to produce statistically valid system quality estimates. In particular, we show that our approach yields valid system quality estimates when either (i) a model trained to predict human ratings using persona ratings and source data observed under sampling bias, or (ii) a reweighting model that corrects for sampling bias is of sufficient quality. We validate our framework theoretically and via a novel Persona Simulation Framework (PSF) designed to systematically manipulate persona quality and the degree of evaluation sampling bias present in source data. Our work provides a principled foundation for combining imperfect persona ratings with human ratings observed under sampling bias to obtain valid system quality estimates.

K. J. Kevin Feng, Tzu-Sheng Kuo, Quan Ze et al.

As LLMs gain adoption in high-stakes domains like mental health, domain experts are increasingly consulted to provide input into policies governing their behavior. From an observation of 19 policymaking workshops with 9 experts over 15 weeks, we identified opportunities to better support rapid experimentation, feedback, and iteration for collaborative policy design processes. We present PolicyPad, an interactive system that facilitates the emerging practice of LLM policy prototyping by drawing from established UX prototyping practices, including heuristic evaluation and storyboarding. Using PolicyPad, policy designers can collaborate on drafting a policy in real time while independently testing policy-informed model behavior with usage scenarios. We evaluate PolicyPad through workshops with 8 groups of 22 domain experts in mental health and law, finding that PolicyPad enhanced collaborative dynamics during policy design, enabled tight feedback loops, and led to novel policy contributions. Overall, our work paves participatory paths for advancing AI alignment and safety.

Luke Guerdan, Amanda Coston, Kenneth Holstein et al.

Predictive models are often introduced to decision-making tasks under the rationale that they improve performance over an existing decision-making policy. However, it is challenging to compare predictive performance against an existing decision-making policy that is generally under-specified and dependent on unobservable factors. These sources of uncertainty are often addressed in practice by making strong assumptions about the data-generating mechanism. In this work, we propose a method to compare the predictive performance of decision policies under a variety of modern identification approaches from the causal inference and off-policy evaluation literatures (e.g., instrumental variable, marginal sensitivity model, proximal variable). Key to our method is the insight that there are regions of uncertainty that we can safely ignore in the policy comparison. We develop a practical approach for finite-sample estimation of regret intervals under no assumptions on the parametric form of the status quo policy. We verify our framework theoretically and via synthetic data experiments. We conclude with a real-world application using our framework to support a pre-deployment evaluation of a proposed modification to a healthcare enrollment policy.

Donghoon Shin, Sachin Grover, Kenneth Holstein et al.

Explainable AI (XAI) is a promising means of supporting human-AI collaborations for high-stakes visual detection tasks, such as damage detection tasks from satellite imageries, as fully-automated approaches are unlikely to be perfectly safe and reliable. However, most existing XAI techniques are not informed by the understandings of task-specific needs of humans for explanations. Thus, we took a first step toward understanding what forms of XAI humans require in damage detection tasks. We conducted an online crowdsourced study to understand how people explain their own assessments, when evaluating the severity of building damage based on satellite imagery. Through the study with 60 crowdworkers, we surfaced six major strategies that humans utilize to explain their visual damage assessments. We present implications of our findings for the design of XAI methods for such visual detection contexts, and discuss opportunities for future research.

Hong Shen, Alicia DeVos, Motahhare Eslami et al.

A growing body of literature has proposed formal approaches to audit algorithmic systems for biased and harmful behaviors. While formal auditing approaches have been greatly impactful, they often suffer major blindspots, with critical issues surfacing only in the context of everyday use once systems are deployed. Recent years have seen many cases in which everyday users of algorithmic systems detect and raise awareness about harmful behaviors that they encounter in the course of their everyday interactions with these systems. However, to date little academic attention has been granted to these bottom-up, user-driven auditing processes. In this paper, we propose and explore the concept of everyday algorithm auditing, a process in which users detect, understand, and interrogate problematic machine behaviors via their day-to-day interactions with algorithmic systems. We argue that everyday users are powerful in surfacing problematic machine behaviors that may elude detection via more centrally-organized forms of auditing, regardless of users' knowledge about the underlying algorithms. We analyze several real-world cases of everyday algorithm auditing, drawing lessons from these cases for the design of future platforms and tools that facilitate such auditing behaviors. Finally, we discuss work that lies ahead, toward bridging the gaps between formal auditing approaches and the organic auditing behaviors that emerge in everyday use of algorithmic systems.

Kenneth Holstein, Shayan Doroudi

The development of educational AI (AIEd) systems has often been motivated by their potential to promote educational equity and reduce achievement gaps across different groups of learners -- for example, by scaling up the benefits of one-on-one human tutoring to a broader audience, or by filling gaps in existing educational services. Given these noble intentions, why might AIEd systems have inequitable impacts in practice? In this chapter, we discuss four lenses that can be used to examine how and why AIEd systems risk amplifying existing inequities. Building from these lenses, we then outline possible paths towards more equitable futures for AIEd, while highlighting debates surrounding each proposal. In doing so, we hope to provoke new conversations around the design of equitable AIEd, and to push ongoing conversations in the field forward.

Kenneth Holstein, Vincent Aleven

Recent work has explored how complementary strengths of humans and artificial intelligence (AI) systems might be productively combined. However, successful forms of human-AI partnership have rarely been demonstrated in real-world settings. We present the iterative design and evaluation of Lumilo, smart glasses that help teachers help their students in AI-supported classrooms by presenting real-time analytics about students' learning, metacognition, and behavior. Results from a field study conducted in K-12 classrooms indicate that students learn more when teachers and AI tutors work together during class. We discuss implications of this research for the design of human-AI partnerships. We argue for more participatory approaches to research and design in this area, in which practitioners and other stakeholders are deeply, meaningfully involved throughout the process. Furthermore, we advocate for theory-building and for principled approaches to the study of human-AI decision-making in real-world contexts.

Pengcheng An, Kenneth Holstein, Bernice d'Anjou et al.

Recently, the HCI community has seen increased interest in the design of teaching augmentation (TA): tools that extend and complement teachers' pedagogical abilities during ongoing classroom activities. Examples of TA systems are emerging across multiple disciplines, taking various forms: e.g., ambient displays, wearables, or learning analytics dashboards. However, these diverse examples have not been analyzed together to derive more fundamental insights into the design of teaching augmentation. Addressing this opportunity, we broadly synthesize existing cases to propose the TA framework. Our framework specifies a rich design space in five dimensions, to support the design and analysis of teaching augmentation. We contextualize the framework using existing designs cases, to surface underlying design trade-offs: for example, balancing actionability of presented information with teachers' needs for professional autonomy, or balancing unobtrusiveness with informativeness in the design of TA systems. Applying the TA framework, we identify opportunities for future research and design.

Kenneth Holstein, Jennifer Wortman Vaughan, Hal Daumé et al.

The potential for machine learning (ML) systems to amplify social inequities and unfairness is receiving increasing popular and academic attention. A surge of recent work has focused on the development of algorithmic tools to assess and mitigate such unfairness. If these tools are to have a positive impact on industry practice, however, it is crucial that their design be informed by an understanding of real-world needs. Through 35 semi-structured interviews and an anonymous survey of 267 ML practitioners, we conduct the first systematic investigation of commercial product teams' challenges and needs for support in developing fairer ML systems. We identify areas of alignment and disconnect between the challenges faced by industry practitioners and solutions proposed in the fair ML research literature. Based on these findings, we highlight directions for future ML and HCI research that will better address industry practitioners' needs.