Emerging evidence shows that writing with AI assistance can change both the views people express and the opinions they hold. Yet, we lack a substantive understanding of behavioral and process-level changes in co-writing with AI that underlie the opinion-shaping power of these tools. We conducted a mixed-methods study, combining retrospective interviews with 19 participants about their co-writing experience with quantitative analysis tracing idea engagement in 1,291 AI co-writing sessions. Our analysis shows that engaging with the AI’s suggestions—reading them and deciding whether to accept them—becomes a central activity, taking away from more traditional processes of ideation and language generation. As writers often do not complete their own ideation before engaging with suggestions, the suggested ideas and opinions seeded directions that writers then elaborated on. At the same time, writers did not notice the AI’s influence and felt in control, as they—in principle—could always edit the final text. We term this shift Reactive Writing: an evaluation-first, suggestion-led writing practice that departs substantially from conventional composing in the presence of AI assistance and is highly vulnerable to AI-induced biases and opinion shifts.

Social media feeds use predictive models to maximize engagement, often misaligning how people consume content with how they wish to. We introduce Bonsai, a system that enables people to build personalized and intentional feeds. Bonsai implements a platform-agnostic framework comprising Planning, Sourcing, Curating, and Ranking modules. This framework allows users to express their intent in natural language and exert fine-grained control over a procedurally transparent feed creation process. We evaluated the system with 15 Bluesky users in a two-phase, multi-week study. We find that participants successfully used our system to discover new content, filter out irrelevant or toxic posts, and disentangle engagement from intent, but curating intentional feeds required more effort than they are used to. Simultaneously, users sought system transparency mechanisms to effectively use (and trust) intentional, personalized feeds. Overall, our work highlights intentional feedbuilding as a viable path beyond engagement-based optimization.

CHI ’25, Yokohama, Japan

This paper examines the tensions between neighborhood gentrification and community surveillance posts on Nextdoor, a hyperlocal social media platform for neighborhoods. We created a privacy-preserving pipeline to gather research data from public Nextdoor posts in Atlanta, Georgia and filtered these to a dataset of 1,537 community surveillance posts. We developed a qualitative codebook to label observed patterns of community surveillance, and deploy a large language model to tag these posts at scale. Ultimately, we present an extensible and empirically-tested typology of the modes of community surveillance that occur on hyperlocal platforms. We find a complex relationship between community surveillance posts and neighborhood gentrification, which indicates that publicly disclosing information about perceived outsiders, especially for petty crimes, is most prevalent in gentrifying neighborhoods. Our empirical evidence inform critical perspectives which posit that community surveillance on platforms like Nextdoor can exclude and marginalize minoritized populations, particularly in gentrifying neighborhoods. Our findings carry broader implications for hyperlocal social platforms and their potential to amplify and exacerbate social tensions and exclusion.

Given large language models’ (LLMs) increasing integration into workplace software, it is important to examine how biases in the models may impact workers. For example, stylistic biases in the language suggested by LLMs may cause feelings of alienation and result in increased labor for individuals or groups whose style does not match. We examine how such writer-style bias impacts inclusion, control, and ownership over the work when co-writing with LLMs. In an online experiment, participants wrote hypothetical job promotion requests using either hesitant or self-assured auto-complete suggestions from an LLM and reported their subsequent perceptions. We found that the style of the AI model did not impact perceived inclusion. However, individuals with higher perceived inclusion did perceive greater agency and ownership, an effect more strongly impacting participants of minoritized genders. Feelings of inclusion mitigated a loss of control and agency when accepting more AI suggestions.

With the steady closure of local newspapers, news consumers increasingly turn to community forums and neighborhood apps to fill the information void. This study investigates how local online groups are perceived relative to more traditional local news outlets, and compares the benefits provided by each information source. Based on prior theoretical contributions, we develop a framework for measuring the benefits of local information on individual-level pro-community attitudes (attachment, knowledge, and civic attitudes.) In a field experiment (N=170), we asked frequent Facebook users living in four U.S. cities to start following local news pages or local online groups on Facebook for one month, and compared their perceptions of source quality and changes in pro-community attitudes. We find that posts from local news pages are perceived to be of higher quality than posts from local online groups. However, following local news pages or local groups did not lead to significant changes in pro-community attitudes during our study period. We discuss implications for the future study of local news in a changing media ecology.

COVID-19 has been a sustained and global crisis with a strong continual impact on daily life. Staying accurately informed about COVID-19 has been key to personal and communal safety, especially for essential workers---individuals whose jobs have required them to go into work throughout the pandemic---as their employment has exposed them to higher risks of contracting the virus. Through 14 semi-structured interviews, we explore how essential workers across industries navigated the COVID-19 information landscape to get up-to-date information in the early months of the pandemic. We find that essential workers living through a sustained crisis have a broad set of information needs. We summarize these needs in a framework that centers 1) fulfilling job requirements, 2) assessing personal risk, and 3) keeping up with crisis news coverage. Our findings also show that the sustained nature of COVID-19 crisis coverage led essential workers to experience breaking points and develop coping strategies. Additionally, we show how workplace communications may act as a mediating force in this process: lack of adequate information in the workplace caused workers to struggle with navigating a contested information landscape, while consistent updates and information exchanges at work could ease the stress of information overload. Our findings extend the crisis informatics field by providing contextual knowledge about the information needs of essential workers during a sustained crisis.

Structured datasets are difficult to keep up-to-date since the underlying facts evolve over time; curated data about business financials, organizational hierarchies, or drug interactions are constantly changing. Drafty is a platform that enlists visitors of an editable dataset to become ``user-editors'' to help solve this problem. It records and analyzes user-editors' within-page interactions to construct user interest profiles, creating a cyclical feedback mechanism that enables Drafty to target requests for specific corrections from user-editors. To validate the automatically generated user interest profiles, we surveyed participants who performed self-created tasks with Drafty and found their user interest score was 3.2 higher on data they were interested in versus data they had no interest in. Next, a 7-month live experiment compared the efficacy of user-editor corrections depending on whether they were asked to review data that matched their interests. Our findings suggest that user-editors are approximately 3 times more likely to provide accurate corrections for data matching their interest profiles, and about 2 times more likely to provide corrections in the first place.