arXiv subjects

Alice Qian

Publications and source records attributed to Alice Qian.

6 recordsLinked to original sources

arXiv2026

On the Framing Effect of AI Red-Teaming Instructions: Evidence from Prior Exercises and Controlled Simulations

AI red-teaming, adversarially probing generative AI systems to surface harms, has become a central evaluation method. However, prior work has highlighted that red-teaming is a fairly open-ended task, laden with myriad design choices that can shape its findings. In this work, we focus on the instructions used to conduct red-teaming exercises and hypothesize that they can significantly affect outcomes in a manner analogous to the framing effect in psychology, where the format in which information is presented greatly impacts how people respond. To understand whether and why instructions impact red-teaming outcomes, first, we analyze the range of ways in which instructions have historically been formulated, reviewing 228 instances of prior work. Second, based on takeaways from this review, we run controlled simulations that hold parameters such as the risk domain, the attacker, and the model under investigation fixed while varying only the instructions. From five real instruction templates used by industry and academic organizations, we derive 45 instruction frames across three risk domains and three levels of risk granularity, generating 100 variants per frame, and deploy each against five open-weights target models, for 22,500 attacks. Using a generalized linear mixed model, we find that certain instruction properties significantly predict attack success rates: the lowest and highest levels of risk granularity both outperform intermediate specificity, as do short and long instructions relative to medium-length ones. The presence of examples is associated with higher attack success rates, while supplying explicit harm definitions yields lower ones. As alignment evaluations become increasingly automated, our results suggest that humans continue to shape red-teaming outcomes substantially through the instructions they provide to scaffold the activity.

cs.CY↗

arXiv2026

Beyond the Single Turn: Reframing Refusals as Dynamic Experiences Embedded in the Context of Mental Health Support Interactions with LLMs

Content Warning: This paper contains participant quotes and discussions related to mental health challenges, emotional distress, and suicidal ideation. Large language models (LLMs) are increasingly used for mental health support, yet the model safeguards -- particularly refusals to engage with sensitive content -- remain poorly understood from the perspectives of users and mental health professionals (MHPs) and have been reported to cause real-world harms. This paper presents findings from a sequential mixed-methods study examining how LLM refusals are experienced and interpreted in mental health support interactions. Through surveys (N=53) and in-depth interviews (N=16) with individuals using LLMs for mental health support and MHPs, we reveal that refusals are not isolated, single-turn system behaviors but rather constitute dynamic, multi-phase experiences: pre-refusal expectation formation, refusal triggering and encounter, refusal message framing, resource referral provision, and post-refusal outcomes. We contribute a multi-phase framework for evaluating refusals beyond binary policy compliance accuracy and design recommendations for future refusal mechanisms. These findings suggest that understanding LLM refusals requires moving beyond single-turn interactions toward recognizing them as holistic experiences embedded within users' support-seeking trajectories and the broader LLM design pipeline.

cs.HC↗

arXiv2026

Towards Apples to Apples for AI Evaluations: From Real-World Use Cases to Evaluation Scenarios

AI measurement science has a wide variety of methodologies and measurements for comparing AI systems, resulting in what often appear to be "apples-to-oranges" comparisons across AI evaluations. To move toward "apples-to-apples" comparisons in real-world AI evaluations, this work advocates for methodological transparency in evaluation scenarios, operational grounding, and human-centered design (HCD) principles. We propose a repeatable process for transforming high-level use cases to detailed scenarios by eliciting use cases from subject matter experts (SMEs) via a structured AI Use Case Worksheet with six key elements: use case, sector, user (direct and indirect), intended outcomes, expected impacts (positive and negative), and KPIs and metrics. We demonstrate utility of the worksheet and process in the U.S. financial services sector. This paper reports on example high-level AI use cases identified by financial services sector SMEs: cyber defense enablement, developer productivity, financial crime aggregation, suspicious activity report (SAR) filing, credit memo generation, and internal call center support. These AI use cases provided are illustrative of the process and not exhaustive. Central to our work is a three-stage expansion pipeline combining LLM prompting with human reviews to generate 107 scenarios from those use cases elicited from SMEs. This process integrates iterative human reviews at every juncture to ensure operational grounding: for scenario titles and descriptions; for core scenario elements like users, benefits and risks, and metrics; and for scenario narratives and evaluation objectives. Human checkpoints ensure scenarios remain reflective of real-world usage and human needs. We describe a validation rubric to assess scenario quality. By defining key scenario components, this work supports a more consistent and meaningful paradigm for human-centered AI evaluations.

cs.HC↗

arXiv2026

Worker Discretion Advised: Co-designing Risk Disclosure in Crowdsourced Responsible AI (RAI) Content Work

Responsible AI (RAI) content work, such as annotation, moderation, or red teaming for AI safety, often exposes crowd workers to potentially harmful content. While prior work has underscored the importance of communicating well-being risk to employed content moderators, designing effective disclosure mechanisms for crowd workers while balancing worker protection with the needs of task designers and platforms remains largely unexamined. To address this gap, we conducted individual co-design sessions with 15 task designers, 11 crowdworkers, and 3 platform representatives. We investigated task designer preferences for support in disclosing tasks, worker preferences for receiving risk disclosure warnings, and how platform representatives envision their role in shaping risk disclosure practices. We identify design tensions and map the sociotechnical tradeoffs that shape disclosure practices. We contribute design recommendations and feature concepts for risk disclosure mechanisms in the context of RAI content work.

cs.HC↗

arXiv2026

Locating Risk: Task Designers and the Challenge of Risk Disclosure in RAI Content Work

As AI systems are increasingly tested and deployed in open-ended and high-stakes domains, crowdworkers are often tasked with responsible AI (RAI) content work. These tasks include labeling violent content, moderating disturbing text, or simulating harmful behavior for red teaming exercises to shape AI system behaviors. While prior research efforts have highlighted the risks to worker well-being associated with RAI content work, far less attention has been paid to how these risks are communicated to workers by task designers or individuals who design and post RAI tasks. Existing transparency frameworks and guidelines, such as model cards, datasheets, and crowdworksheets, focus on documenting model information and dataset collection processes, but they overlook an important aspect of disclosing well-being risks to workers. In the absence of standard workflows or clear guidance, the consistent application of content warnings, consent flows, or other forms of well-being risk disclosure remains unclear. This study investigates how task designers approach risk disclosure in crowdsourced RAI tasks. Drawing on interviews with 23 task designers across academic and industry sectors, we examine how well-being risk is recognized, interpreted, and communicated in practice. Our findings highlight the need to support task designers in identifying and communicating risks not only to support crowdworker well-being but also to strengthen the ethical integrity and technical efficacy of AI development pipelines.

cs.HC↗

arXiv2025

Exploring Opportunities to Support Novice Visual Artists' Inspiration and Ideation with Generative AI

Recent generative AI advances present new possibilities for supporting visual art creation, but how such promise might assist novice artists during early-stage processes requires investigation. How novices adopt or resist these tools can shift the relationship between the art community and generative systems. We interviewed 13 artists to uncover needs in key dimensions during early stages of creation: (1) quicker and better access to references, (2) visualizations of reference combinations, (3) external artistic feedback, and (4) personalized support to learn new techniques and styles. Mapping such needs to state-of-the-art open-sourced advances, we developed a set of six interactive prototypes to expose emerging capabilities to novice artists. Afterward, we conducted co-design workshops with 13 novice visual artists through which artists articulated requirements and tensions for artist-centered AI development. Our work reveals opportunities to design novice-targeted tools that foreground artists' needs, offering alternative visions for generative AI to serve visual creativity.

cs.HC↗