arXiv subjects

K. J. Kevin Feng

Publications and source records attributed to K. J. Kevin Feng.

18 recordsLinked to original sources

arXiv2026

Hedwig: Dynamic Autonomy for Coding Agents Under Local Oversight

Despite coding agents' advances in handling increasingly complex tasks, their continued tendency to introduce unintended edits, subtle bugs, and scope drift that slip past code review means developers must still decide how much autonomy to grant them. However, existing approaches for setting an agent's level of autonomy, such as static permission settings or instruction files, cannot account for how developers' preferences for agent autonomy can shift across tasks and over time. We conducted a formative survey with 21 software engineers who use coding agents and found that they experience frustration with calibrating autonomy and have evolving preferences for level of oversight. Building on these insights, we present Hedwig, a CLI coding agent that dynamically adjusts its autonomy level based on developer-agent interactions across sessions. Rather than operating on a global, fixed autonomy configuration, Hedwig learns an evolving set of behavioral guidelines from developer decisions and feedback, reducing friction on work for which the agent has earned trust, while tightening oversight when the agent operates outside familiar territory. Hedwig demonstrates the potential of a new paradigm where agents intelligently adapt their level of autonomy based on user trust through active, longitudinal collaboration.

cs.HC↗

arXiv2026

PolicyPad: Collaborative Prototyping of LLM Policies

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 expert-informed paths for advancing AI alignment and safety.

cs.HC↗

arXiv2026

Cocoa: Co-Planning and Co-Execution with AI Agents

As AI agents take on increasingly long-running tasks involving sophisticated planning and execution, there is a corresponding need for novel interaction designs that enable deeper human-agent collaboration. However, most prior works leverage human interaction to fix "autonomous" workflows that have yet to become fully autonomous or rigidly treat planning and execution as separate stages. Based on a formative study with 9 researchers using AI to support their work, we propose a design that affords greater flexibility in collaboration, so that users can 1) delegate agency to the user or agent via a collaborative plan where individual steps can be assigned; and 2) interleave planning and execution so that plans can adjust after partial execution. We introduce Cocoa, a system that takes design inspiration from computational notebooks to support complex research tasks. A lab study (n=16) found that Cocoa enabled steerability without sacrificing ease-of-use, and a week-long field deployment (n=7) showed how researchers collaborated with Cocoa to accomplish real-world tasks.

cs.HC↗

arXiv2025

On the Regulatory Potential of User Interfaces for AI Agent Governance

AI agents that take actions in their environment autonomously over extended time horizons require robust governance interventions to curb their potentially consequential risks. Prior proposals for governing AI agents primarily target system-level safeguards (e.g., prompt injection monitors) or agent infrastructure (e.g., agent IDs). In this work, we explore a complementary approach: regulating user interfaces of AI agents as a way of enforcing transparency and behavioral requirements that then demand changes at the system and/or infrastructure levels. Specifically, we analyze 22 existing agentic systems to identify UI elements that play key roles in human-agent interaction and communication. We then synthesize those elements into six high-level interaction design patterns that hold regulatory potential (e.g., requiring agent memory to be editable). We conclude with policy recommendations based on our analysis. Our work exposes a new surface for regulatory action that supplements previous proposals for practical AI agent governance.

cs.CY↗

arXiv2025

Democratic AI is Possible. The Democracy Levels Framework Shows How It Might Work

This position paper argues that effectively "democratizing AI" requires democratic governance and alignment of AI, and that this is particularly valuable for decisions with systemic societal impacts. Initial steps -- such as Meta's Community Forums and Anthropic's Collective Constitutional AI -- have illustrated a promising direction, where democratic processes could be used to meaningfully improve public involvement and trust in critical decisions. To more concretely explore what increasingly democratic AI might look like, we provide a "Democracy Levels" framework and associated tools that: (i) define milestones toward meaningfully democratic AI, which is also crucial for substantively pluralistic, human-centered, participatory, and public-interest AI, (ii) can help guide organizations seeking to increase the legitimacy of their decisions on difficult AI governance and alignment questions, and (iii) support the evaluation of such efforts.

cs.CY↗

arXiv2025

Levels of Autonomy for AI Agents

Autonomy is a double-edged sword for AI agents, simultaneously unlocking transformative possibilities and serious risks. How can agent developers calibrate the appropriate levels of autonomy at which their agents should operate? We argue that an agent's level of autonomy can be treated as a deliberate design decision, separate from its capability and operational environment. In this work, we define five levels of escalating agent autonomy, characterized by the roles a user can take when interacting with an agent: operator, collaborator, consultant, approver, and observer. Within each level, we describe the ways by which a user can exert control over the agent and open questions for how to design the nature of user-agent interaction. We then highlight a potential application of our framework towards AI autonomy certificates to govern agent behavior in single- and multi-agent systems. We conclude by proposing early ideas for evaluating agents' autonomy. Our work aims to contribute meaningful, practical steps towards responsibly deployed and useful AI agents in the real world.

cs.HC↗

arXiv2025

IdeaSynth: Iterative Research Idea Development Through Evolving and Composing Idea Facets with Literature-Grounded Feedback

Research ideation involves broad exploring and deep refining ideas. Both require deep engagement with literature. Existing tools focus primarily on idea broad generation, yet offer little support for iterative specification, refinement, and evaluation needed to further develop initial ideas. To bridge this gap, we introduce IdeaSynth, a research idea development system that uses LLMs to provide literature-grounded feedback for articulating research problems, solutions, evaluations, and contributions. IdeaSynth represents these idea facets as nodes on a canvas, and allow researchers to iteratively refine them by creating and exploring variations and composing them. Our lab study (N=20) showed that participants, while using IdeaSynth, explored more alternative ideas and expanded initial ideas with more details compared to a strong LLM-based baseline. Our deployment study (N=7) demonstrated that participants effectively used IdeaSynth for real-world research projects at various ideation stages from developing initial ideas to revising framings of mature manuscripts, highlighting the possibilities to adopt IdeaSynth in researcher's workflows.

cs.HC↗

arXiv2025

Interactive Reasoning: Visualizing and Controlling Chain-of-Thought Reasoning in Large Language Models

The output quality of large language models (LLMs) can be improved via "reasoning": generating segments of chain-of-thought (CoT) content to further condition the model prior to producing user-facing output. While these chains contain valuable information, they are verbose and lack explicit organization, making them tedious to review. Moreover, they lack opportunities for user feedback, such as to remove unwanted considerations, add desired ones, or clarify unclear assumptions. We introduce Interactive Reasoning, an interaction design that visualizes chain-of-thought outputs as a hierarchy of topics and enables user review and modification. We implement interactive reasoning in Hippo, a prototype for AI-assisted decision making in the face of uncertain trade-offs. In a user study with 16 participants, we find that interactive reasoning in Hippo allows users to quickly identify and interrupt erroneous generations, efficiently steer the model towards customized responses, and better understand both model reasoning and model outputs. Our work contributes to a new paradigm that incorporates user oversight into LLM reasoning processes.

cs.HC↗

arXiv2025

Policy Prototyping for LLMs: Pluralistic Alignment via Interactive and Collaborative Policymaking

Emerging efforts in AI alignment seek to broaden participation in shaping model behavior by eliciting and integrating collective input into a policy for model finetuning. While pluralistic, these processes are often linear and do not allow participating stakeholders to confirm whether potential outcomes of their contributions are indeed consistent with their intentions. Design prototyping has long advocated for rapid iteration using tight feedback loops of ideation, experimentation, and evaluation to mitigate these issues. We thus propose policy prototyping for LLMs, a new process that draws inspiration from prototyping practices to enable stakeholders to collaboratively and interactively draft LLM policies. Through learnings from a real-world LLM policymaking initiative at an industrial AI lab, we motivate our approach and characterize policy prototyping with four guiding principles. Because policy prototyping emphasizes a contrasting set of priorities compared to previous approaches, we envision our approach to be a valuable addition to the methodological repertoire for collaborative, pluralistic alignment.

cs.HC↗

arXiv2025

Canvil: Designerly Adaptation for LLM-Powered User Experiences

Advancements in large language models (LLMs) are sparking a proliferation of LLM-powered user experiences (UX). In product teams, designers often craft UX to meet user needs, but it is unclear how they engage with LLMs as a novel design material. Through a formative study with 12 designers, we find that designers seek a translational process that enables design requirements to shape and be shaped by LLM behavior, motivating a need for designerly adaptation to facilitate this translation. We then built Canvil, a Figma widget that operationalizes designerly adaptation. We used Canvil as a probe to study designerly adaptation in a group-based design study (6 groups, N=17), finding that designers constructively iterated on both adaptation approaches and interface designs to enhance end-user interaction with LLMs. Furthermore, designers identified promising collaborative workflows for designerly adaptation. Our work opens new avenues for processes and tools that foreground designers' human-centered expertise when developing LLM-powered applications.

cs.HC↗

arXiv2024

SPICA: Retrieving Scenarios for Pluralistic In-Context Alignment

When different groups' values differ, one approach to model alignment is to steer models at inference time towards each group's preferences. However, techniques like in-context learning only consider similarity when drawing few-shot examples and not cross-group differences in values. We propose SPICA, a framework that accounts for group-level differences during in-context example retrieval. SPICA introduces three designs: scenario banks, group-informed retrieval metrics, and in-context alignment prompts. From an evaluation of SPICA on an alignment task collecting inputs from four demographic groups ($n = 544$), our metrics retrieve in-context examples that more closely match observed preferences, with the best prompt configuration using multiple contrastive responses to demonstrate examples. In an end-to-end evaluation ($n = 120$), we observe that SPICA is higher rated than similarity-based retrieval, with groups seeing up to a +0.16 point improvement on a 5 point scale. Additionally, gains from SPICA were more uniform, with all groups benefiting from alignment rather than only some. Finally, we find that while a group-agnostic approach can align to aggregated values, it is not most suited for divergent groups.

cs.CL↗

arXiv2024

(A)I Am Not a Lawyer, But...: Engaging Legal Experts towards Responsible LLM Policies for Legal Advice

Large language models (LLMs) are increasingly capable of providing users with advice in a wide range of professional domains, including legal advice. However, relying on LLMs for legal queries raises concerns due to the significant expertise required and the potential real-world consequences of the advice. To explore \textit{when} and \textit{why} LLMs should or should not provide advice to users, we conducted workshops with 20 legal experts using methods inspired by case-based reasoning. The provided realistic queries ("cases") allowed experts to examine granular, situation-specific concerns and overarching technical and legal constraints, producing a concrete set of contextual considerations for LLM developers. By synthesizing the factors that impacted LLM response appropriateness, we present a 4-dimension framework: (1) User attributes and behaviors, (2) Nature of queries, (3) AI capabilities, and (4) Social impacts. We share experts' recommendations for LLM response strategies, which center around helping users identify `right questions to ask' and relevant information rather than providing definitive legal judgments. Our findings reveal novel legal considerations, such as unauthorized practice of law, confidentiality, and liability for inaccurate advice, that have been overlooked in the literature. The case-based deliberation method enabled us to elicit fine-grained, practice-informed insights that surpass those from de-contextualized surveys or speculative principles. These findings underscore the applicability of our method for translating domain-specific professional knowledge and practices into policies that can guide LLM behavior in a more responsible direction.

cs.CY↗

arXiv2024

Mapping the Design Space of Teachable Social Media Feed Experiences

Social media feeds are deeply personal spaces that reflect individual values and preferences. However, top-down, platform-wide content algorithms can reduce users' sense of agency and fail to account for nuanced experiences and values. Drawing on the paradigm of interactive machine teaching (IMT), an interaction framework for non-expert algorithmic adaptation, we map out a design space for teachable social media feed experiences to empower agential, personalized feed curation. To do so, we conducted a think-aloud study (N=24) featuring four social media platforms -- Instagram, Mastodon, TikTok, and Twitter -- to understand key signals users leveraged to determine the value of a post in their feed. We synthesized users' signals into taxonomies that, when combined with user interviews, inform five design principles that extend IMT into the social media setting. We finally embodied our principles into three feed designs that we present as sensitizing concepts for teachable feed experiences moving forward.

cs.HC↗

arXiv2023

Case Repositories: Towards Case-Based Reasoning for AI Alignment

Case studies commonly form the pedagogical backbone in law, ethics, and many other domains that face complex and ambiguous societal questions informed by human values. Similar complexities and ambiguities arise when we consider how AI should be aligned in practice: when faced with vast quantities of diverse (and sometimes conflicting) values from different individuals and communities, with whose values is AI to align, and how should AI do so? We propose a complementary approach to constitutional AI alignment, grounded in ideas from case-based reasoning (CBR), that focuses on the construction of policies through judgments on a set of cases. We present a process to assemble such a case repository by: 1) gathering a set of ``seed'' cases -- questions one may ask an AI system -- in a particular domain, 2) eliciting domain-specific key dimensions for cases through workshops with domain experts, 3) using LLMs to generate variations of cases not seen in the wild, and 4) engaging with the public to judge and improve cases. We then discuss how such a case repository could assist in AI alignment, both through directly acting as precedents to ground acceptable behaviors, and as a medium for individuals and communities to engage in moral reasoning around AI.

cs.AI↗

arXiv2023

Examining the Impact of Provenance-Enabled Media on Trust and Accuracy Perceptions

In recent years, industry leaders and researchers have proposed to use technical provenance standards to address visual misinformation spread through digitally altered media. By adding immutable and secure provenance information such as authorship and edit date to media metadata, social media users could potentially better assess the validity of the media they encounter. However, it is unclear how end users would respond to provenance information, or how to best design provenance indicators to be understandable to laypeople. We conducted an online experiment with 595 participants from the US and UK to investigate how provenance information altered users' accuracy perceptions and trust in visual content shared on social media. We found that provenance information often lowered trust and caused users to doubt deceptive media, particularly when it revealed that the media was composited. We additionally tested conditions where the provenance information itself was shown to be incomplete or invalid, and found that these states have a significant impact on participants' accuracy perceptions and trust in media, leading them, in some cases, to disbelieve honest media. Our findings show that provenance, although enlightening, is still not a concept well-understood by users, who confuse media credibility with the orthogonal (albeit related) concept of provenance credibility. We discuss how design choices may contribute to provenance (mis)understanding, and conclude with implications for usable provenance systems, including clearer interfaces and user education.

cs.HC↗

arXiv2023

How Do UX Practitioners Communicate AI as a Design Material? Artifacts, Conceptions, and Propositions

UX practitioners (UXPs) face novel challenges when working with and communicating artificial intelligence (AI) as a design material. We explore how UXPs communicate AI concepts when given hands-on experience training and experimenting with AI models. To do so, we conducted a task-based design study with 27 UXPs in which they prototyped and created a design presentation for a AI-enabled interface while having access to a simple AI model training tool. Through analyzing UXPs' design presentations and post-activity interviews, we found that although UXPs struggled to clearly communicate some AI concepts, tinkering with AI broadened common ground when communicating with technical stakeholders. UXPs also identified key risks and benefits of AI in their designs, and proposed concrete next steps for both UX and AI work. We conclude with a sensitizing concept and recommendations for design and AI tools to enhance multi-stakeholder communication and collaboration when crafting human-centered AI experiences.

cs.HC↗

arXiv2023

Understanding Collaborative Practices and Tools of Professional UX Practitioners in Software Organizations

User experience (UX) has undergone a revolution in collaborative practices, due to tools that enable quick feedback and continuous collaboration with a varied team across a design's lifecycle. However, it is unclear how this shift in collaboration has been received in professional UX practice, and whether new pain points have arisen. To this end, we conducted a survey (N=114) with UX practitioners at software organizations based in the U.S. to better understand their collaborative practices and tools used throughout the design process. We found that while an increase in collaborative activity enhanced many aspects of UX work, some long-standing challenges -- such as handing off designs to developers -- still persist. Moreover, we observed new challenges emerging from activities enabled by collaborative tools such as design system management. Based on our findings, we discuss how UX practices can improve collaboration moving forward and provide concrete design implications for collaborative UX tools.

cs.HC↗

arXiv2023

Addressing UX Practitioners' Challenges in Designing ML Applications: an Interactive Machine Learning Approach

UX practitioners face novel challenges when designing user interfaces for machine learning (ML)-enabled applications. Interactive ML paradigms, like AutoML and interactive machine teaching, lower the barrier for non-expert end users to create, understand, and use ML models, but their application to UX practice is largely unstudied. We conducted a task-based design study with 27 UX practitioners where we asked them to propose a proof-of-concept design for a new ML-enabled application. During the task, our participants were given opportunities to create, test, and modify ML models as part of their workflows. Through a qualitative analysis of our post-task interview, we found that direct, interactive experimentation with ML allowed UX practitioners to tie ML capabilities and underlying data to user goals, compose affordances to enhance end-user interactions with ML, and identify ML-related ethical risks and challenges. We discuss our findings in the context of previously established human-AI guidelines. We also identify some limitations of interactive ML in UX processes and propose research-informed machine teaching as a supplement to future design tools alongside interactive ML.

cs.HC↗