CS&E Excels at 2026 ACM SIGCHI Conference with 13 Accepted Papers

Authors: Anna Martin-Boyle, William Humphreys, Martha Brown, Cara Leckey, Harmanpreet Kaur

Scholars everywhere recognize that LLM outputs are prone to errors – but what are these errors? We asked scientists to evaluate LLM answers to questions about their own papers, then catalogued every mistake they found. The result is a schema of 20 error types across 7 categories: from hallucinated citations and invented equations, to synthesis failures across multiple papers, and answers that are technically correct but dangerously incomplete. Standard automated benchmarks miss most of this; you need real experts evaluating real outputs to see how LLMs actually fail in scholarly work.

Authors: Nuredin Ali Abdelkadir, Tianling Yang, Shivani Kapania, Kauna Ibrahim Malgwi, Fasica Berhane Gebrekidan, Adio-Adet Dinika, Elaine O. Nsoesie, Milagros Miceli, Stevie Chancellor

Recently documented the terrible mental health conditions of African content moderators working for Meta and TikTok contractors. In a survey of 134 moderators, and interviews with 15 of them, we found that these moderators experience significantly worse mental health outcomes than their counterparts in other regions. Former moderators show even higher distress levels, suggesting long-term harm. The root causes appear to be systemic, tied to the labor conditions, poor pay, and poor mental health support of these workers.

Authors: Xianhao Carton Liu, Difan Jia, Tongyu Nie, Evan Suma Rosenberg, Victoria Interrante, Chen Zhu-Tian

Artificial Intelligence (AI) and indoor sensing increasingly support decision-making in spatial environments. However, traditional visualization methods impose a substantial mental workload when viewers translate this digital information into real-world spaces, leading to inappropriate reliance on AI. Embedded visualizations in Augmented Reality (AR), by integrating information into physical environments, may reduce this workload and foster more appropriate reliance on AI. To assess this, we conducted an empirical study (N = 32) comparing an AR embedded visualization (X-ray) and 2D Minimap in AI-assisted, time-critical spatial target selection tasks. Surprisingly, evidence shows that the embedded visualization led to greater inappropriate reliance on AI, primarily as over-reliance, due to factors like perceptual challenges, visual proximity illusions, and highly realistic visual representations. Nonetheless, the embedded visualization showed benefits in spatial mapping. We conclude by discussing empirical insights, design implications, and directions for future research on human-AI collaborative decision in AR.

Authors: Shichen Liang, Alvina Salim, Svetlana Yarosh, Amanda Johnson, Ji Youn Shin

Family caregivers of patients with chronic conditions take on many responsibilities, often at the expense of their own health. Previous studies have examined technological supports, such as information access and online social groups. While these approaches are promising, they primarily focus on easing the load of a single caregiver. Even when interventions are designed to support collaborative caregiving, they often overlook which tasks are shareable and what influences caregivers’ decisions to delegate. To address these gaps, we conducted activity-based qualitative interviews with 10 family caregivers of patients who underwent life-threatening medical procedures. Our findings show that collaboration occurs when tasks are shareable in practice, including tangible and logistical help, and when caregivers judge that the benefits of receiving support outweigh the work of seeking help. We propose that technologies should support caregivers in determining when collaboration is feasible by recognizing task types, addressing the trade-offs embedded in activation costs, and enabling both proactive and reactive assistance as situations evolve.

Authors: Mo Houtti, Moyan Zhou, Daniel Runningen, Surabhi Sunil, Leor Porat, Harmanpreet Kaur, Loren Terveen, Stevie Chancellor

We built an AI agent to improve inclusion through better feedback exchange. The agent uses the Induced Hypocrisy procedure, a well-grounded technique from social psychology. This procedure can prompt people to make behavior changes by highlighting hypocrisies in their own behaviors. We show that it works in the lab with an experimental study, but organizational factors made people use the tool for personal reflection, NOT improving meeting inclusion.

Authors: Anna Martin-Boyle, Cara Leckey, Martha Brown, Harmanpreet Kaur

We built a system called PaperTrail to help researchers verify whether LLM-generated answers to scholarly questions are actually supported in the source papers. Instead of just showing source links or citations, PaperTrail breaks both the LLM's answer and the source documents into specific claims and evidence, showing you what's supported, what's missing, and what might be made up. The system made researchers more skeptical of the AI, BUT that skepticism didn't change how much they relied on the LLM. We outline a trust-behavior gap: turns out, knowing an LLM might be wrong isn't enough to make people act on that suspicion. 

Authors: Ashlee Milton, Dan Runningen, Loren Terveen, Harmanpreet Kaur, Stevie Chancellor

We found that social media AI algorithms, like those that support TikTok FYP, are impossible for people to control when they need to avoid harmful content. Through workshops with people, researchers have created a process model to describe why and how it affects people. This paper gives us a clearer mechanism and explanation for the process of how social media AI algorithms cannot be controlled, and potential design solutions to these problems.

Authors: Matthew Brehmer, Maxime Cordeil, Christophe Hurter, Takayuki Itoh, Wolfgang Büschel, Mahmood Jasim, Arnaud Prouzeau, David Saffo, Lyn Bartram, Sheelagh Carpendale, Chen Zhu-Tian, Andrew Cunningham, Tim Dwyer, Samuel Huron, Masahiko Itoh, Alark Joshi, Kiyoshi Kiyokawa, Hideaki Kuzuoka, Bongshin Lee, Gabriela Molina León, Harald Reiterer, Bektur Ryskeldiev, Jonathan Schwabish, Brian A. Smith, Yasuyuki Sumi, Ryo Suzuki, Anthony Tang, Yalong Yang, Jian Zhao

We characterize 16 challenges faced by those investigating and developing remote and synchronous collaborative experiences around visualization. Our work reflects the perspectives and prior research efforts of an international group of 29 experts from across human-computer interaction and visualization sub-communities. The challenges are anchored around five collaborative activities that exhibit a centrality of visualization and multimodal communication. These activities include exploratory data analysis, creative ideation, visualization-rich presentations, joint decision making grounded in data, and real-time data monitoring. The challenges also reflect the changing dynamics of these activities in the face of recent advances in extended reality (XR) and artificial intelligence (AI). As an organizing scheme for future research at the intersection of visualization and computer-supported cooperative work, we align the challenges with a sequence of four sets of research and development activities: technological choices, social factors, AI assistance, and evaluation.

Authors: Kelly Wang, Ashlee Milton, Leah Namisa Rosenbloom, Erika Melder, Ada Lerner, Michael Ann DeVito

Online platforms have become a key avenue for forming new relationships, especially for queer individuals. However, some individuals, such as those in asexual and aromantic communities (A-Spec), seek forms of relationships that trouble existing frameworks assumed by online platforms, such as dating apps. To investigate A-Spec needs, we conducted an 8-week ARC study with 38 A-Spec participants who have used online platforms for developing relationships. Participants described a mismatch between the design of dating apps and their approach to building relationships, suggesting platform design that combines affordances of dating apps and other social platforms. We thus outline a ``process-oriented'' paradigm for relationship-building platforms inspired by community design suggestions, supporting participants' process of first establishing a low-stakes relationship and then co-constructing its properties. We also argue for a ``pluralized'' approach to defining identity and relationship in the design of online systems, upsetting default assumptions surrounding any given label.

Authors: Jasmine C Foriest, Leah Ajmani, Munmun De Choudhury

Intimate partner violence (IPV) is defined as “abuse or aggression that occurs in a romantic relationship." IPV survivors face barriers when help-seeking, such as epistemic injustice -- secondary victimization from dismissal and indifference when disclosing, misdirection, or inappropriate interventions. Survivors may leverage generative AI to make sensitive disclosures and access hermeneutic resources. However, these tools mediate outcomes for IPV survivors through novel manifestations of epistemic injustice. Using mixed-methods, we investigated hermeneutic resource provision by large-language models (LLMs). We evaluated LLM responses to IPV disclosures on three axes: hermeneutic resource provision, readability, and risk. Prompts were derived from a content analysis of IPV and generative AI discussions in 5 abuse subreddits. We contribute a taxonomy of 7 uses of generative AI in the experience of IPV, empirical illustration of epistemic inequity, and considerations for evaluating epistemic harm in generative AI. Content Warning: This study contains descriptions of abuse and violence.

Authors: Shivani Kapania, Tianling Yang, Nuredin Ali, Morgan Klaus Scheuerman, Milagros Miceli, Alex S Taylor, Sarah E Fox

The rapid expansion of the AI industry relies heavily on the production, verification, and maintenance of data, otherwise known as "data work". Companies outsource and offshore this work through global AI supply chains that operate under exploitative conditions. Drawing on semi-structured interviews with Kenyan data workers across platforms and BPOs, this paper examines how such conditions take shape and persist. We argue that workers are caught within a regime of entrapment, a system of interconnected mechanisms that make it difficult for workers to leave or improve their positions. These mechanisms include the push to invest in the promise of ‘AI’ jobs, the use of precarious contracts to govern workers, the capture of regulatory institutions, and the exploitation of global labor arbitrage. Using complementary lenses of neoliberal governmentality, precarity, and supply chain capitalism, we analyze why labor mobilization in this sector remains uniquely constrained. We conclude by outlining an orientation for research and scholarly practice that can support workers' organizing efforts and contest the structural conditions sustaining this regime.

Authors: Chunggi Lee, Hayato Saiki, Tica Lin, EIJI IKEDA, Kenji Suzuki, Chen Zhu-Tian, Hanspeter Pfister

We present ViSTAR, a Virtual Skill Training system in AR that supports self-guided basketball skill practice, with feedback on balance, posture, and timing. From a formative study with basketball players and coaches, the system addresses three challenges: understanding skills, identifying errors, and correcting mistakes. ViSTAR follows the Behavioral Skills Training (BST) framework—instruction, modeling, rehearsal, and feedback. It provides feedback through visual overlays, rhythm and timing cues, and an AI-powered coaching agent using 3D motion reconstruction. We generate verbal feedback by analyzing spatio-temporal joint data and mapping features to natural-language coaching cues via a Large Language Model (LLM). A key novelty is this feedback generation: motion features become concise coaching insights. In two studies (N=16), participants generally preferred our AI-generated feedback to coach feedback and reported that ViSTAR helped them notice posture and balance issues and refine movements beyond self-observation.

Authors: Sharon Heung, Claire Florence Weizenegger, Mo Houtti, Sunny Consolvo, Patrick Gage Kelley, Tara Matthews, Renee Shelby, Kurt Thomas, Ashley Marie Walker

A growing body of qualitative research has identified contextual risk factors that elevate people’s chances of experiencing digital-safety attacks. However, the lack of quantitative data on the population-level distribution of these risk factors prevents policymakers and tech companies from developing targeted, evidence-based interventions to improve digital safety. To address this gap, we surveyed 5,001 adults in the United States to analyze: (1) the frequency of and relationship between digital-safety attacks (e.g., scams, harassment, account hacking), and (2) how these attacks align with 10 contextual risk factors. Nearly half of our respondents identify as resource constrained, which significantly correlates with higher likelihood of experiencing four common attacks. We also present qualitative insights to expand our understanding of the factors beyond the existing literature (e.g., “prominence” included high-visibility roles in local communities). This study provides the first large-scale quantitative analysis correlating digital-safety attacks with contextual risk factors and demographics.