Survival | Contentious Times | Grand Challenges: UMSI Research Roundup
Monday, 11/24/2025
By Noor HindiUniversity of Michigan School of Information faculty and PhD students are creating and sharing knowledge that helps build a better world. Here are some of their recent publications.
Publications
Web scraping for research: Legal, ethical, institutional, and scientific considerations
Big Data and Society, November 2025
Megan A Brown, Andrew Gruen, Gabe Maldoff, Solomon Messing, Zeve Sanderson, Michael Zimmer
Scientists across disciplines often use data from the internet to conduct research, generating valuable insights about human behavior. However, as generative artificial intelligence relying on massive text corpora becomes increasingly valuable, platforms have greatly restricted access to data through official channels. As a result, researchers will likely engage in more web scraping to collect data, introducing new challenges and concerns for researchers. This paper proposes a comprehensive framework for web scraping in social science research for U.S.-based researchers, examining the legal, ethical, institutional, and scientific factors that we recommend researchers consider when scraping the web. We present an overview of the current regulatory environment impacting when and how researchers can access, collect, store, and share data via scraping. We then provide researchers with recommendations to conduct scraping in a scientifically legitimate and ethical manner. We aim to equip researchers with the relevant information to mitigate risks and maximize the impact of their research amid this evolving data access landscape.
Relational Infrastructures of Survival: How Social Dependencies Enable Alternative Internets in Cuba
University of Pittsburgh Press, November 2025
Over the past several decades, Cubans have developed multiple collective strategies to navigate extended periods of constraint and precarity, from resource shortages to a global pandemic. With the increasing presence of internet technologies in Havana, digital media has become entangled in these processes, resulting in overlapping internet ecosystems supported by human relationships. This article explores the social nature of internet engagements in Havana through the lens of relational infrastructure—the people, relationships, and social practices that Cubans rely on to sustain overlapping internet ecosystems as they adapt and endure social, economic, and political pressures. Drawing on ethnographic data, I describe how people in Havana achieve their goals by stitching together the digital, the physical, and the social. Looking at social engagements through this lens reveals power dynamics and structural inequalities that challenge assumptions regarding the positive impact of internet technologies.
“Unnecessarily cumbersome”: Researchers' Opinions on Restricted Data Access Systems
Proceedings of the Association for Information Science and Technology, October 2025
Megan A Brown, Andrea Thomer, Libby Hemphill
Research data archives use restricted data access protocols to manage access to sensitive data. However, restricted data access systems can be cumbersome for researchers to engage in data reuse, as the systems frequently implemented introduce friction into the research process. We fielded a survey of 481 data reusers at the Inter-university Consortium for Political and Social Research (ICPSR) in 2020 about restricted data access systems. We found that 80% of respondents would be more likely to reuse data if restricted data access applications were made faster and easier. Additionally, most researchers indicated they believe that the security of research data is very important. However, researchers disagreed on the appropriate set of mechanisms to ensure that research data remains secure, especially discounting interventions that introduce friction to accessing data. These findings present challenges for archives in implementing restricted data access systems that balance protecting research subjects with encouraging data reuse.
Scholarly Productivity in Contentious Times: Future Considerations for Early Career Information Scholars
Proceedings of the Association for Information Science and Technology, October 2025
Megan Threats, Rebecca D. Frank, Angela D. R. Smith, Katrina Fenlon, Andrea Thomer
Recent policy changes and sweeping cuts to federal agencies in the United States (US) pose a significant threat to information scholars and practitioners in the US and elsewhere who benefit from the funding, services, programming, and support made possible by federal agencies like the Institute of Museum and Library Services (IMLS), National Institutes of Health (NIH), and the National Science Foundation (NSF). The termination of research grants, the deletion of public federal data sets, and mass layoffs across the federal sector in the US have left many early career scholars concerned about disruptions to their research and scholarly productivity. These disruptions have the potential to impact scholars around the world who, for example, rely on data that is under threat, or who collaborate with researchers based at institutions in the US. This panel will discuss future considerations for these actions' impact on information scholars, practitioners, and their communities. We will present strategies for fostering scholarly productivity through scholarly collaboration, data sharing and reuse, and information resilience. We aim to foster an open discussion with panelists and audience members to explore additional avenues and strategies that early career information scholars may pursue to navigate these challenges.
Personal Archival Practices: Broadening Our Understanding of Archival Stewardship
Proceedings of the Association for Library and Information Science Education (ALISE) Conference, October 2025
Archival stewardship is a useful concept in understanding an archivist’s management of records and collections. Without the practical application of stewardship, archives would be less accessible and intelligible to the public, far less physically protected, and disconnected from the community from which it may stem. This work remains invaluable in the archival field, but it is both inadequate in capturing and frequently leaves out the care, labor, and intellectual contributions of community members who are stewards of their own archives prior to donating them to institutions. By better understanding their archival stewardship work, this paper broadens the archival field’s definition of stewardship to include unique, community-specific contributions, its findings impacting core theoretical concerns such as intellectual contributions to history and archival silences. Using a feminist epistemological framework, this paper takes seriously knowledge produced by a community and thus broadens and strengthens the model of archival stewardship as we know it.
A Critical Reflection on Designing and Evaluating a Personalized Self-Care Intervention for Care Partners: Lessons from a Randomized Controlled Trial
10th International Workshop on Mental Health and Well-being: From Research to Practice in Mental Healthcare, October 2025
Rongqi Bei, Christopher Graves, Mark W. Newman, Predrag Klasnja, Jennifer Miner, Sung Won Choi, Angelle M. Sander, Noelle E. Carlozzi
Care partners often face significant mental strain, but are rarely able to prioritize their own well-being. To support this vulnerable group, we designed the CareQOL app, an adaptive intervention combining daily self-monitoring of caregiver strain, depression, and anxiety with Fitbit-tracked health behaviors to deliver personalized push notifications targeting self-care. We evaluated this app in a six-month randomized controlled trial with care partners (N=254). Despite high compliance, adding personalized notifications to self-monitoring did not significantly improve participants’ mental health or other health-related quality of life outcomes, though participants who rated the app more usable demonstrated a greater likelihood of improvement. This prompted reflection on intervention design and study process, facilitated by follow-up interviews (N=36). We found that personalization based solely on passive sensing and self-reports did not always match participants’ dynamic needs and preferences, while dashboard visualizations without contextual details hindered in-depth self-reflection. Additionally, extended study durations made it hard for participants to recall experiences and left systems vulnerable to technical breakdowns. We offer practical recommendations for research at the intersection of ubiquitous computing and mental healthcare: enable user-initiated personalization through lightweight interactions, support contextual annotations in self-monitoring dashboards, adopt data-prompted interview approaches, and implement batch enrollment with technical fallback strategies.
Making Bodies: Assumptions in the Design and Validation of Motion Capture Technology
Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society, October 2025
Mona Sloane, Abigail Jacobs, Emanuel Moss
Motion capture technologies are increasingly being incorporated into key aspects of social life, but embed potentially harmful assumptions. We examine the structural factors that allow for assumptions to solidify in motion capture systems and propose a matrix that works as a structured guide for motion capture developers to identify and document the assumptions baked into the system they are working on. To motivate novel conversations and future work in this area, we conclude with discussing the application of this matrix as well as its limitations.
Signals in the Noise: Decoding Unexpected Engagement Patterns on Twitter
Proceedings of the ACM on Human-Computer Interaction, October 2025
Yulin Yu, Houming Chen, Daniel Romero, Paramveer S. Dhillon
Social media platforms offer users multiple ways to engage with content-likes, retweets, and comments-creating a complex signaling system within the attention economy. While previous research has examined factors driving overall engagement, less is known about why certain tweets receive unexpectedly high levels of one type of engagement relative to others. Drawing on Signaling Theory and Attention Economy Theory, we investigate these unexpected engagement patterns on Twitter. The social media platform has been renamed to 'X,' however our study data is from the time-period prior to the name change., developing an ''unexpectedness quotient'' to quantify deviations from predicted engagement levels. Our analysis of over 600,000 tweets reveals distinct patterns in how content characteristics influence unexpected engagement. News, politics, and business tweets receive more retweets and comments than expected, suggesting users prioritize sharing and discussing informational content. In contrast, games and sports-related topics garner unexpected likes and comments, indicating higher emotional investment in these domains. The relationship between content attributes and engagement types follows clear patterns: subjective tweets attract more likes while objective tweets receive more retweets, and longer, complex tweets with URLs unexpectedly receive more retweets. These findings demonstrate how users employ different engagement types as signals of varying strength based on content characteristics, and how certain content types more effectively compete for attention in the social media ecosystem. Our results offer valuable insights for content creators optimizing engagement strategies, platform designers facilitating meaningful interactions, and researchers studying online social behavior.
Mapping the Landscape, Measuring the Gap:Qualitative Methods Reporting in Information Science Research
Proceedings of the Association for Information Science and Technology, October 2025
Rebecca Frank, Adam Kriesberg
We examined qualitative methods reporting in Information Science research by analyzing ASIS&T conferencepapers (2018-2022) and comparing findings with journal publishing guidelines. Our study of 117 papers usingexclusively qualitative methods revealed significant gaps in methodological documentation. While 78.6% of papersinvolved human subjects research (primarily interviews), only 28.3% mentioned IRB approval. Similarly, 66.7%failed to describe analytical tools used. Journal publishing guidelines across the field showed inconsistentrequirements for qualitative research reporting, with some mandating IRB disclosure while others provided minimaldirection. The prevalent use of passive voice in methods sections often obscured critical information about dataproducers and collection processes. These findings demonstrate a need for more standardized reporting guidelinesfor qualitative research in Information Science. We recommend that ASIS&T publishing venues require authors toprovide, at minimum: data production year(s), clear identification of data producers, persistent identifiers whenavailable, and IRB approval status for human subjects research. These measures could enhance transparency andfacilitate better understanding of qualitative research practices in the field.
“Trying to Piece It Together”: Exploring Accessible Error Detection in Emerging Privacy Techniques With Blind People
ASSETS ‘25: Proceedings of the 27th International ACM SIGACCESS Conference on Computers and Accessibility, October 2025
Rahaf Alharbi, Angela D Cheong, Jaylin Herskovitz, Robin N. Brewer, Sarita Schoenebeck
Blind people use visual assistance technologies (VAT) to access visual information, yet VAT can expose blind people to privacy risks. Prior HCI research has studied and built AI-enabled obfuscation techniques to detect and remove private content. However, blind people cannot easily spot errors in obfuscation tools. Our paper explores how assessment descriptors, brief visual attributes of objects, may enable blind people to find errors. By conducting interviews and focus groups with blind participants, we found that certain assessment descriptors (color, dimensions, distance) are inadequate to support blind people. Instead, participants discussed assessment descriptors that better reflect their sensemaking process, such as describing multiple objects in a particular space. Expanding the scope of accessible verification beyond assessment descriptors, participants called for greater transparency on how AI-enabled privacy techniques are developed and emphasized the need to co-create training materials on using AI-enabled privacy techniques. Building from our findings and disability studies scholarship, our paper examines how sighted bias could produce assessment descriptors that neglect the needs of blind people and analyzes how participants’ preferred assessment descriptors contrast with existing standards of visual description. Lastly, we offer design directions to push for greater transparency in VAT and obfuscation tools.
Last-mile Work of Infrastructures: Collective Coordination of Early COVID-19 Vaccine Distribution for Older Adults
Computer Supported Cooperative Work (CSCW): The Journal of Collaborative Computing and Work Practices, September 2025
Sam A. Ankenbauer, Alex Jiahong Lu
Handled by a constellation of public and private institutions, the COVID-19 vaccine infrastructure in the U.S. was largely fragmented when it first emerged. By looking into the early phases of the COVID-19 vaccination distribution among older adults in the U.S. state of Florida, this work attends to the “last-mile logistics” of vaccine infrastructure—the final leg of distribution, where vaccines meet their intended end consumers. Drawing on in-depth interviews with older adults and the volunteers who supported them in securing vaccination appointments, we illustrate the complexities of the last mile and the challenges faced by older adults in navigating distribution. We then demonstrate how volunteers and older adults came together as an emerging assemblage to collectively perform what we term “last-mile work”—a specific form of infrastructuring undertaken in the last mile to resolve localized logistic or operational challenges within a larger infrastructural arrangement. This paper introduces last-mile work as a critical yet under-theorized form of infrastructuring that is indispensable to infrastructure’s everyday function. We argue that last-mile work offers an important analytical lens and empirical site for future infrastructural investigations in CSCW research and design. It creates new entry points to rethink the assumed relationships between infrastructuring work and temporal breakdowns, while also surfacing critical questions of labor, power, and (in)visibility within the last mile of infrastructures.
Designing Hands-Free Technology to Support Real-Time Patient Data Collection and Documentation for Emergency Care Settings
International Journal of Human-Computer Interaction, September 2025
Zhan Zhang, Enze Bai, Yincao Xu, Aram Stepanian, Sun Young Park
Using handheld electronic health record (EHR) devices to collect and document patient data poses significant challenges in dynamic, time-critical, and hands-busy medical settings. Prior research has proposed wearable technologies, such as smart glass, to enable hands-free clinical documentation. Building on this, we conducted a two-year, user-centered study to iteratively design and evaluate a smart glass application for enhancing real-time clinical documentation in settings like Emergency Medical Services (EMS). Our findings provide key design insights for addressing EMS documentation challenges through smart glass technology, as well as potential barriers for successful adoption. We conclude the paper by discussing the implications of these findings for developing smart glass to support documentation in fast-paced medical environments.
A dynamic Bayesian network approach to modeling engagement and walking behavior: insights from a yearlong micro-randomized trial (Heartsteps II)
Issue: Understanding and Predicting the Temporal Dynamics of Health and Wellbeing, September 2025
Steven A. De La Torre, Mohamed El Mistiri, Karine Tung, Eric Hekler, Predrag Klasnja, Misha Pavel, Daniel E. Rivera, Donna Spruijt-Metz, Benjamin Marlin
Introduction: Mobile health (mHealth) technologies such as wearable activity trackers (e.g. Fitbit) and digital applications (apps), can support behavior change in real-world contexts. Since effectiveness is dependent, in part, on participants’ engagement with the digital technology (e.g. app page views) and the intervention components (e.g. anti-sedentary messages), there is a need for modeling approaches that support the investigation of engagement in digital interventions and the refinement of dynamic theories of behavior change.
Methods: Dynamic Bayesian Networks (DBN) were used to model the idiographic (individual) dynamic relationships between a participant’s daily app engagement (page views), walking behavior, and intervention messages, accounting for context (e.g. temperature), and psychological variables (e.g. perceived restedness and perceived busyness). Additionally, we explored differences in the resulting DBN models between participants of Hispanic/Latino and non-Hispanic/ Latino White backgrounds.
Results: Data from 10 participants in the HeartSteps II study (n = 5 Hispanic/Latinos and n = 5 non-Hispanic/Latino Whites) was used. Across participants (100%, n = 10), there was a strong positive effect of the number of messages/prompts received on their daily app page views with a predicted increase range of 12.84 (12.19–13.57) to 25.84 (24.28–27.59) app page views per day per message received. Among the majority of Hispanic/Latino participants (n = 4/5, 80%), there was a strong positive relationship between daily app page views and walking behavior with predictions ranging from a mean of 6.70 (6.37–7.05) to 10.93 (10.14– 11.78) steps per minute of Fitbit wear time per app page view. Both groups showed idiographic differences in the effects of temperature and perceived busyness on walking behavior.
Conclusion: The results demonstrate the benefits of DBNs to model the daily-level idiographic behavioral dynamics of engagement in digital intervention studies. This approach can be leveraged to support the refinement of dynamic theories of behavior change and improving personalized mHealth intervention strategies.
Exploring #Diasporawars on Black Twitter
Social Media + Society, September 2025
Tyler Musgrave, Yuning Ye, Kentaro Toyama, Sarita Schoenebeck, Megan Threats
Black Twitter, now operating on X (formerly Twitter), is a crucial online platform that shapes cultural production, political activism, and educational exchange within the global Black community. This study broadens the scope to examine the global influence of Black Twitter, with a focus on the hashtag #diasporawars. This hashtag serves as a lens through which we can observe the tensions and interactions across the global Black community. Black Twitter’s significance extends beyond the United States, deeply embedded in the historical and cultural contexts of Blackness, which inform global conversations on identity. By employing both quantitative and qualitative research methods to analyze #diasporawars, this study aims to shed light on the complexities of global Blackness and how social media platforms contribute to shaping these identities and connections. Our findings reveal that #diasporawars reflects broader dynamics within the global Black community, highlighting how platforms like X both facilitate positive engagement and exacerbate conflicts. This research underscores the multifaceted nature of Black digital spaces, illustrating how they serve as arenas for collaboration and contention, influenced by diverse experiences and perspectives within the global Black diaspora.
Daily Dietary Sodium Intake Among Clinical Trial Participants Recruited From a University Health System or a Federally Qualified Health Center: Secondary Analysis of Baseline Participant Characteristics
JMIR Cardio, September 2025
Gabriella V Rubick, Michael P Dorsch, Scott L Hummel, Tanima Basu, Evan Luff, Kimberly Warden, Michael Giacalone, Sarah Bailey, Mark W Newman, Lesli E Skolarus, Brahmajee K Nallamothu, Jessica R Golbus
Background: Efforts to improve diversity in clinical trials often prioritize recruitment based on broad demographic factors. This approach may overlook the influence of community context and health-related social needs on health behaviors, including sodium intake, a key modifiable risk factor for hypertension and cardiovascular disease.
Objective: This study aims to assess the impact of enrollment site, sociodemographic factors, and health-related social needs on baseline dietary sodium intake among participants in a mobile health clinical trial aimed at lowering blood pressure.
Methods: The myBPmyLife study is a prospective, randomized controlled trial evaluating a mobile health intervention to lower blood pressure through increased physical activity and lower sodium food choices. Participants with hypertension were recruited from a university health system and a federally qualified health center (FQHC). All participants completed a validated sodium screener at enrollment. Sociodemographic data and health-related social needs were self-reported. Univariable and multivariable linear regression models were used to evaluate the associations between sodium intake and participant characteristics. This analysis presents a cross-sectional examination of the baseline characteristics of participants enrolled in the myBPmyLife study.
Results: Among 600 included participants, 96 (16.0%) were from the FQHC. Mean age was 60.1 (SD 13.5) years; 48.2% (289/600) were women, and 13.0% (78/600) were Black. FQHC participants were significantly younger (mean age 47.9, SD 11.1 vs 62.5, SD 12.7 years), more likely to be Black (43/96, 44.8% vs 35/504, 6.9%), and 8.5 times more likely to have difficulty paying for their health-related social needs. Mean baseline sodium intake was 3082.3 (SD 1072.5) mg/day, with 85.5% (513/600) of participants exceeding the World Health Organization’s recommended daily sodium limit. Baseline sodium intake was significantly higher for FQHC participants (mean difference 381.1, SD 1064.2 mg/d; 95% CI 84.5-677.7; P=.01), men (mean difference 543.9, SD 1038.3 mg/d; 95% CI 377.3-710.5; P<.001), Black participants (mean difference 442.5, SD 1043.4 mg/d; 95% CI 119.7-765.3; P=.008) and those with difficulty affording basic needs (mean difference 338.1, SD 1066.7 mg/d; 95% CI 95.2-581.0; P=.02). Sodium intake was lower in older participants (−196.4 mg/d per 10 years; 95% CI −258.0 to −134.9; P<.001). In a multivariable analysis, age, gender, and race remained independently associated with sodium intake, while differences by site and health-related social needs were not statistically significant.
Conclusion: Differences in sodium intake were observed across sociodemographic groups. While the enrollment site was not independently associated with sodium intake after adjustment, it played a role in shaping the participant population, evidenced by the differences in demographics and health-related social needs among participants based on enrollment site. These findings underscore the importance of recruiting from distinct clinical settings to capture a range of contextual factors that influence health behaviors. Clinical trials aiming for representativeness should consider both individual- and community-level factors during recruitment to more accurately inform interventions and health outcomes.
Counter-surveying apartheid-era forced removals in South Africa: a spatial approach to archival social justice
Archival Science, September 2025
Siddique Motala, Tlotliso Mokomane, David A. Wallace
This paper describes and reports on applications of the counter-surveying methodology that reads together geomatics (surveying and mapping) and archival social justice (activation of archives to serve social justice outcomes and counter injustice) in relation to two sites of forced removal: one in District Six in Cape Town and the other in Die Vlakte in Stellenbosch, South Africa. Archives from multiple sources are activated to supplement the recording of the stories of ex-residents who experienced forced removal. Ex-residents further engage through a process of annotating maps from their neighbourhood. In combination this documentation enriches and deepens knowledge about these forced removals. Counter-surveying works to ensure that these demolitions and their ongoing impacts on lives are documented so that their histories and existence are not rendered invisible for future generations. Ex-residents, most of whom are facing their final decades, hold the last remnants of living memory connected to these sites. Counter-surveying provides the opportunity for ex-residents to revisit and recall in the wake of community demolition. This work is motivated by an activist approach and memory for justice ethic that focuses explicitly on land and memory. We close this paper with key findings and invitation to others to take counter-surveying as a praxis and methodology that can be meaningfully applied to other sites of forced removal, working with both the original inhabitants or their descendants who were likewise impacted.
Multi-Click: Cross-Tab Web Automation via Action Generalization
UIST '25: Proceedings of the 38th Annual ACM Symposium on User Interface Software and Technology, September 2025
Jiacheng Zhang, Jiawen Li, Maryam Arab, Steve Oney
Repetitive actions are a common and frustrating part of using the web. Prior work has proposed automating repetitive actions with natural language descriptions, demonstrations, and pseudocode. However, these approaches introduce abstractions that can be difficult to write, evaluate, and fit within web workflows. We describe a new approach, Multi-Click, for simultaneously performing the same action (e.g., clicking or typing) across multiple pages while maintaining the immediacy and understandability of direct manipulation. Users can intuitively select groups of analogous elements within or across windows/tabs (e.g., equivalent elements in different instantiations of a template) and interact with these elements as if each simultaneously had keyboard or cursor focus (e.g., one click propagates to multiple targets). Multi-Click introduces algorithms for identifying analogous elements from structural and visual attributes; techniques for intuitively selecting and visualizing targets; and uses interactive data grids to manage variation in text entry and retrieval tasks.
Privacy Equilibrium: Balancing Privacy Needs in Dynamic Multi-User Augmented Reality Scenarios
UIST '25: Proceedings of the 38th Annual ACM Symposium on User Interface Software and Technology, September 2025
Shwetha Rajaram, Jiasi Chen, Michael Nebeling
As augmented reality (AR) glasses become more widely used in public settings, a key challenge is meeting the privacy needs of multiple AR users and bystanders in a fine-grained manner. To enable this, we present a conceptual framework for Privacy Equilibrium–balancing user experience (UX) and privacy between all individuals in a shared space. The framework applies constrained optimization to compute AR sensing policies that grant or restrict permissions to maximize UX while minimizing privacy risks (e.g., capturing bystanders or sensitive environmental data). We instantiate this framework in a simulation and analysis toolkit to holistically evaluate different optimization strategies and visualize tradeoffs between UX and privacy. Through application scenarios, we demonstrate the flexibility of our optimization approach to minimize these tradeoffs across conflicting user needs and privacy preferences. Walkthrough evaluations with AR and security & privacy researchers highlight the potential of our framework and toolkit to inform future privacy-mediating techniques for AR.
Interactions Between Workers and Automated Guided Vehicles: Impact of eHMI Design
Proceedings of the Human Factors and Ergonomics Society Annual Meeting, September 2025
Doo Won Han, Shreyas Bhat, Shaoze Yang, Justin Smith, Al Salour, Terra Stroup, Paul Pridham, X. Jessie Yang
As manufacturing facilities integrate Autonomated Guided Vehicles (AGVs) to improve workflow efficiency, enhancing humanAGV interaction remains critical for workplace safety. While prior research has focused on vehicle and pedestrian motion prediction, effective interaction requires two-way communication, where the AGVs clearly convey intentions to the workers to enhance safety. This study investigates the impact of external Human-Machine Interface (eHMI) integrated with a predictive model on AGV-worker interaction. We designed LED light strip patterns to convey intentions and selected optimal designs through an online survey. We deployed three types of AGVs in a virtual reality (VR) environment: Control, Prediction, and eHMI + Prediction. Participants completed tasks while interacting with AGVs, followed by subjective assessments of trust, perceived safety, perceived performance, and understandability. A one-way repeated measures ANOVA revealed a significant improvement in perceived safety from eHMI + Prediction condition compared to the Control condition, suggesting that explicit communication via eHMI enhances perceived safety in AGV interactions.
Grand Challenges in Human-Centered Privacy
IEEE, August 2025
Ruba Abu-Salma, Pauline Anthonysamy, Zinaida Benenson, Benjamin Berens, Kovila P. L. Coopamootoo, Andreas Gutmann, Adam Jenkins, Sameer Patil, Sören Preibusch, Florian Schaub, William Seymour, Jose Such, Mohammad Tahaei, Aybars Tuncdogan, Max Van Kleek, Daricia Wilkinson
We report the most salient themes and future directions for human-centered privacy that emerged from the discussions at the Future of Human-Centred Privacy event, which brought together leading international experts from academia, industry, and government to discuss grand challenges in the field.
Between court orders and platform policies: understanding law enforcement and meta interactions in addressing non-consensual image disclosure abuse
SOUPS ‘25: Proceedings of the Twenty-First USENIX Conference on Usable Privacy and Security, August 2025
Non-Consensual Image Disclosure Abuse (NCIDA) occurs when one person posts, or threatens to post, sensitive images of another person online with the intent to extort, humiliate, or harm. Though much is known about NCIDAs, almost nothing is known about how law enforcement agencies (LEAs) work with social media companies to address them, especially outside the West. Through discussions with Pakistani law enforcement, and legal experts, and analysis of LEA requests submitted to Meta platforms, we find that platforms are reasonably proactive in responding to NCIDA-related requests. However, their decisions are seem to be heavily influenced by their universal content-moderation policies, which are determined by Western norms that prioritize sexually explicit content but neglect content considered sensitive in other cultures. Our findings contribute a nuanced understanding of the communication between LEAs and social media companies in combating NCIDA, and lead to recommendations for platforms and government policy in mitigating NCIDA.
Pre-prints, Working Papers, Articles, Workshops and Talks
Framing Unionization on Facebook: Communication around Representation Elections in the United States
arXiv, October 2025
Arianna Pera, Veronica Jude, Ceren Budak, Luca Maria Aiello
Digital media have become central to how labor unions communicate, organize, and sustain collective action. Yet little is known about how unions' online discourse relates to concrete outcomes such as representation elections. This study addresses the gap by combining National Labor Relations Board (NLRB) election data with 158k Facebook posts published by U.S. labor unions between 2015 and 2024. We focused on five discourse frames widely recognized in labor and social movement communication research: diagnostic (identifying problems), prognostic (proposing solutions), motivational (mobilizing action), community (emphasizing solidarity), and engagement (promoting interaction). Using a fine-tuned RoBERTa classifier, we systematically annotated unions' posts and analyzed patterns of frame usage around election events. Our findings showed that diagnostic and community frames dominated union communication overall, but that frame usage varied substantially across organizations. In election cases that unions won, communication leading up to the vote showed an increased use of diagnostic, prognostic, and community frames, followed by a reduction in prognostic and motivational framing after the event--patterns consistent with strategic preparation. By contrast, in lost election cases unions showed little adjustment in their communication, suggesting an absence of tailored communication strategies. By examining variation in message-level framing, the study highlights how communication strategies adapt to organizational contexts, contributing open tools and data and complementing prior research in understanding digital communication of unions and social movements.
Active Measuring in Reinforcement Learning With Delayed Negative Effects
arXiv, October 2025
Daiqi Gao, Ziping Xu, Aseel Rawashdeh, Predrag Klasnja, Susan A. Murphy
Measuring states in reinforcement learning (RL) can be costly in real-world settings and may negatively influence future outcomes. We introduce the Actively Observable Markov Decision Process (AOMDP), where an agent not only selects control actions but also decides whether to measure the latent state. The measurement action reveals the true latent state but may have a negative delayed effect on the environment. We show that this reduced uncertainty may provably improve sample efficiency and increase the value of the optimal policy despite these costs. We formulate an AOMDP as a periodic partially observable MDP and propose an online RL algorithm based on belief states. To approximate the belief states, we further propose a sequential Monte Carlo method to jointly approximate the posterior of unknown static environment parameters and unobserved latent states. We evaluate the proposed algorithm in a digital health application, where the agent decides when to deliver digital interventions and when to assess users' health status through surveys.
Big Reasoning with Small Models: Instruction Retrieval at Inference Time
arXiv, October 2025
Kenan Alkiek, David Jurgens, Vinod Vydiswaran
Can we bring large-scale reasoning to local-scale compute? Small language models (SLMs) are increasingly attractive because they run efficiently on local hardware, offering strong privacy, low cost, and reduced environmental impact. Yet they often struggle with tasks that require multi-step reasoning or domain-specific knowledge. We address this limitation through instruction intervention at inference time, where an SLM retrieves structured reasoning procedures rather than generating them from scratch. Our method builds an Instruction Corpus by grouping similar training questions and creating instructions via GPT-5. During inference, the SLM retrieves the most relevant instructions and follows their steps. Unlike retrieval-augmented generation, which retrieves text passages, instruction retrieval gives the model structured guidance for reasoning. We evaluate this framework on MedQA (medical board exams), MMLU Professional Law, and MathQA using models from 3B to 14B parameters without any additional fine-tuning. Instruction retrieval yields consistent gains: 9.4% on MedQA, 7.9% on MMLU Law, and 5.1% on MathQA. Concise instructions outperform longer ones, and the magnitude of improvement depends strongly on model family and intrinsic reasoning ability.
When Collaborative Maintenance Falls Short: The Persistence of Retracted Papers on Wikipedia
arXiv, September 2025
Haohan Shi, Yulin Yu, Daniel M. Romero, Emőke-Ágnes Horvát
Wikipedia serves as a key infrastructure for public access to scientific knowledge, but it faces challenges in maintaining the credibility of cited sources—especially when scientific papers are retracted. This paper investigates how citations to retracted research are handled on English Wikipedia. We construct a novel dataset that integrates Wikipedia revision histories with metadata from Retraction Watch, Crossref, Altmetric, and OpenAlex, identifying 1,181 citations of retracted papers. We find that 71.6% of all citations analyzed are problematic. These are citations added before a paper’s retraction, as well as the citations introduced after retraction without any in-text mention of the paper’s retracted status. Our analysis reveals that these citations persist for a median of over 3.68 years (1,344 days). Through survival analysis, we find that signals of human attention are associated with a faster correction process. Unfortunately, a paper’s established scholarly authority—a higher academic citation count—is associated with a slower time to correction. Our findings highlight how the Wikipedia community supports collaborative maintenance but leaves gaps in citation-level repair. We contribute to CSCW research by advancing our understanding of this sociotechnical vulnerability, which takes the form of a community coordination challenge, and by offering design directions to support citation credibility at scale.
RELATED
Keep up with research from UMSI experts by subscribing to our free research roundup newsletter!