Social media users employ a variety of methods to avoid unwanted content in their personalized feeds. Platforms like Instagram and YouTube offer the ''Not Interested'' feature, allowing users to signal their preference to see less of certain content or similar types. Despite its wide availability on social media platforms, the ''Not Interested'' feature has received little academic attention, leaving a gap in understanding how it is perceived, used, and interpreted. This study investigates (1) the types of content users mark as ''Not Interested'', (2) their expectations about its effects, and (3) how they interpret its outcome. We conducted semi-structured interviews with 28 Instagram users who had used the ''Not Interested'' button more than 12 times over the past year, focusing on their experiences with this feature. Users used the ''Not Interested'' feature to avoid different types of content, ranging from problematic and personally discomforting to unrewarding. ''Not Interested'' feedback was considered unique in that it would remove similar content from their feeds without harming the creator. Users had mixed expectations about whose feed they wished their feedback to have an impact on. After submitting the feedback and observing the changes in the personalized feed, they were often uncertain of how each of their interactions was reflected in the algorithm. We discuss user characteristics that social media platforms need to consider, including diverse motivations related to content avoidance, users' demands to influence a broad range of content curation, and the need for granular control and deeper understanding of personalized algorithms.

Label set construction—deciding on a group of distinct labels—is an essential stage in building a supervised machine learning (ML) application, as a badly designed label set negatively affects subsequent stages, such as training dataset construction, model training, and model deployment. Despite its significance, it is challenging for ML practitioners to come up with a well-defined label set, especially when no external references are available. Through our formative study (n=8), we observed that even with the help of external references or domain experts, ML practitioners still need to go through multiple iterations to gradually improve the label set. In this process, there exist challenges in collecting helpful feedback and utilizing it to make optimal refinement decisions. To support informed refinement, we present DynamicLabels, a system that aims to support a more informed label set-building process with crowd feedback. Crowd workers provide annotations and label suggestions to the ML practitioner’s label set, and the ML practitioner can review the feedback through multi-aspect analysis and refine the label set with crowd-made labels. Through a within-subjects study (n=16) using two datasets, we found that DynamicLabels enables better understanding and exploration of the collected feedback and supports a more structured and flexible refinement process. The crowd feedback helped ML practitioners explore diverse perspectives, spot current weaknesses, and shop from crowd-generated labels. Metrics and label suggestions in DynamicLabels helped in obtaining a high-level overview of the feedback, gaining assurance, and spotting surfacing conflicts and edge cases that could have been overlooked.

Simple up/downvotes, arguably the most widely used reaction design across social media platforms, allow users to efficiently express their opinions and quickly evaluate others’ opinions from aggregated votes. However, such design forces users to project their diverse opinions onto dichotomized reactions and provides limited information to readers on why a comment was up/downvoted. We explore user-generated labels (UGLs) as an alternative reaction design to capture the rich context of user reactions to comments. We conducted a between-subjects study with 218 participants to understand how people use and are influenced by UGLs compared to up/downvotes. Specifically, we examine how UGLs affect users’ ability to express and perceive diverse opinions. Participants generated 234 unique labels on diverse aspects of a comment. Leaving more reactions than participants in the up/downvotes condition, participants reported that the ability to express their opinions improved with UGLs. UGLs also enabled participants to better understand the multifacetedness of public evaluation of a comment.

Online chat is an emerging channel for discussing community problems. It is common practice for communities to assign dedicated moderators to maintain a structured discussion and enhance the problem-solving experience. However, due to the synchronous nature of online chat, moderators face a high managerial overhead in tasks like discussion stage management, opinion summarization, and consensus-building support. To assist moderators with facilitating a structured discussion for community problem-solving, we introduce SolutionChat, a system that (1) visualizes discussion stages and featured opinions and (2) recommends contextually appropriate moderator messages. Results from a controlled lab study (n=55, 12 groups) suggest that participants' perceived discussion trackability was significantly higher with SolutionChat than without. Also, moderators provided better summarization with less effort and better managerial support using system-generated messages with SolutionChat than without. With SolutionChat, we envision untrained moderators to effectively facilitate chat-based discussions of important community matters.

Online instructional videos are ubiquitous, but it is difficult for instructors to gauge learners' experience and their level of comprehension or confusion regarding the lecture video. Moreover, learners watching the videos may become disengaged or fail to reflect and construct their own understanding. This paper explores instructor and learner perceptions of in-video prompting where learners answer reflective questions while watching videos. We conducted two studies with crowd workers to understand the effect of prompting in general, and the effect of different prompting strategies on both learners and instructors. Results show that some learners found prompts to be useful checkpoints for reflection, while others found them distracting. Instructors reported the collected responses to be generally more specific than what they have usually collected. Also, different prompting strategies had different effects on the learning experience and the usefulness of responses as feedback.

Despite recent efforts in opening up government data, developing tools for taxpayers to make sense of extensive and multi-faceted budget data remains an open challenge. In this paper, we present BudgetMap, an issue-driven classification and navigation interface for the budgets of government programs. Our novel issue-driven approach can complement the traditional budget classification system used by government organizations by reflecting time-evolving public interests. BudgetMap elicits the public to tag government programs with social issues by providing two modes of tagging. User-initiated tagging allows people to voluntarily search for programs of interest and classify each program with related social issues, while system-initiated tagging guides people through possible matches of issues and programs via microtasks. BudgetMap then facilitates visual exploration of the tagged budget data. Our evaluation shows that participants' awareness and understanding of budgetary issues increased after using BudgetMap, while they collaboratively identified issue-budget links with quality comparable to expert-generated links.

While a government budget determines how taxpayers' money is allocated to various programs and stakeholders that compete for limited resources, the extensiveness and complexity of the budget and its process hinder taxpayers from understanding the budget information and participating in the public discussion. To engage taxpayers in the public discussion around budgetary issues, we leverage news articles containing budgetary information for design opportunities. We present Factful, a web-based annotative article reading interface that enhances the article with fact-checking support and contextual budgetary information by processing open government data. In our lab study, participants using Factful discussed more critically with more fact-based supporting statements. They built a rich context surrounding the relevant budget facts beyond what was presented in the article. Factful presents a simple yet powerful model for supporting fact-oriented budgetary discussions online by leveraging open government data.