Modern algorithmic recommendation systems seek to engage users through behavioral content-interest matching. While many platforms recommend content based on engagement metrics, others like TikTok deliver interest-based content, resulting in recommendations perceived to be hyper-personalized compared to other platforms. TikTok's robust recommendation engine has led some users to suspect that the algorithm knows users "better than they know themselves," but this is not always true. In this paper, we explore TikTok users' perceptions of recommended content on their For You Page (FYP), specifically calling attention to unwanted recommendations. Through qualitative interviews of 14 current and former TikTok users, we find themes of frustration with recommended content, attempts to rid themselves of unwanted content, and various degrees of success in eschewing such content. We discuss implications in the larger context of folk theorization and contribute concrete tactical and behavioral examples of algorithmic persistence.

The surge in the popularity of text-to-image generators (T2Is) has been matched by extensive research into ensuring fairness and equitable outcomes, with a focus on how they impact society. However, such work has typically focused on globally-experienced identities or centered Western contexts. In this paper, we address interpretations, representations, and stereotypes surrounding a tragically underexplored context in T2I research: caste. We examine how the T2I Stable Diffusion displays people of various castes, and what professions they are depicted as performing. Generating 100 images per prompt, we perform CLIP-cosine similarity comparisons with default depictions of an `Indian person’ by Stable Diffusion, and explore patterns of similarity. Our findings reveal how Stable Diffusion outputs perpetuate systems of `castelessness’, equating Indianness with high-castes and depicting caste-oppressed identities with markers of poverty. In particular, we note the stereotyping and representational harm towards the historically-marginalized Dalits, prominently depicted as living in rural areas and always at protests. Our findings underscore a need for a caste-aware approach towards T2I design, and we conclude with design recommendations.

Though research into text-to-image generators (T2Is) such as Stable Diffusion has demonstrated their amplification of societal biases and potentials to cause harm, such research has primarily relied on computational methods instead of seeking information from real users who experience harm, which is a significant knowledge gap. In this paper, we conduct the largest human subjects study of Stable Diffusion, with a combination of crowdsourced data from 133 crowdworkers and 14 semi-structured interviews across diverse countries and genders. Through a mixed-methods approach of intra-set cosine similarity hierarchies (i.e., comparing multiple Stable Diffusion outputs for the same prompt with each other to examine which result is `closest' to the prompt) and qualitative thematic analysis, we first demonstrate a large disconnect between user expectations for Stable Diffusion outputs with those generated, evidenced by a set of Stable Diffusion renditions of `a Person' providing images far away from such expectations. We then extend this finding of general dissatisfaction into highlighting representational harms caused by Stable Diffusion upon our subjects, especially those with traditionally marginalized identities, subjecting them to incorrect and often dehumanizing stereotypes about their identities. We provide recommendations for a harm-aware approach to (re)design future versions of Stable Diffusion and other T2Is.

Our research investigates the impact of Generative Artificial Intelligence (GAI) models, specifically text-to-image generators (T2Is), on the representation of non-Western cultures, with a focus on Indian contexts. Despite the transformative potential of T2Is in content creation, concerns have arisen regarding biases that may lead to misrepresentations and marginalizations. Through a Non-Western community-centered approach and grounded theory analysis of 5 focus groups from diverse Indian subcultures, we explore how T2I outputs to English input prompts depict Indian culture and its subcultures, uncovering novel representational harms such as exoticism and cultural misappropriation. These findings highlight the urgent need for inclusive and culturally sensitive T2I systems. We propose design guidelines informed by a sociotechnical perspective, contributing to the development of more equitable and representative GAI technologies globally. Our work underscores the necessity of adopting a community-centered approach to comprehend the sociotechnical dynamics of these models, complementing existing work in this space while identifying and addressing the potential negative repercussions and harms that may arise as these models are deployed on a global scale.

In this multicultural age, language translation is one of the most performed tasks, and it is becoming increasingly AI-moderated and automated. As a novel AI system, ChatGPT claims to be proficient in machine translation tasks and in this paper, we put that claim to the test. Specifically, we examine ChatGPT’s accuracy in translating between English and languages that exclusively use gender-neutral pronouns. We center this study around Bengali, the 7th most spoken language globally, but also generalize our findings across five other languages: Farsi, Malay, Tagalog, Thai, and Turkish. We find that ChatGPT perpetuates gender defaults and stereotypes assigned to certain occupations (e.g., man = doctor, woman = nurse) or actions (e.g., woman = cook, man = go to work), as it converts gender-neutral pronouns in languages to ‘he’ or ‘she’. We also observe ChatGPT completely failing to translate the English gender-neutral singular pronoun ‘they’ into equivalent gender-neutral pronouns in other languages, as it produces translations that are incoherent and incorrect. While it does respect and provide appropriately gender-marked versions of Bengali words when prompted with gender information in English, ChatGPT appears to confer a higher respect to men than to women in the same occupation. We conclude that ChatGPT exhibits the same gender biases which have been demonstrated for tools like Google Translate or MS Translator, as we provide recommendations for a human centered approach for future designers of AI systems that perform machine translation to better accommodate such low-resource languages.

Background and context. Concept inventories (CIs) are a widely used tool in STEM education that can help instructors identify specific misconceptions students hold about key concepts. Over the past several years, much research has been published contributing to CIs in computer science education.

Over the past few years, AI bias has become a central concern within design and computing fields. But as the concept of bias has grown in visibility, its meaning and form have become harder to grasp. To help designers realize bias, we take inspiration from textile bias (the skew of woven material) and examine the topic across its myriad forms: visual, textual, and tactile. By introducing a slanted experience of material and therefore of reality, we explore the translation of fraught machine learning algorithms into personal and probing artifacts. In this pictorial, we present nine pieces that materialize complex relationships with machine learning; ground these relationships in the present and the personal; and point to generative ways of engaging with biased systems around us.

Online health communities offer the promise of support benefits to users, in particular because these communities enable users to find peers with similar experiences. Building mutually supportive connections between peers is a key motivation for using online health communities. However, a user's role in a community may influence the formation of peer connections. In this work, we study patterns of peer connections between two structural health roles: patient and non-professional caregiver. We examine user behavior in an online health community---CaringBridge.org---where finding peers is not explicitly supported. This context lets us use social network analysis methods to explore the growth of such connections in the wild and identify users' peer communication preferences. We investigated how connections between peers were initiated, finding that initiations are more likely between two authors who have the same role and who are close within the broader communication network. Relationships---patterns of repeated interactions---are also more likely to form and be more interactive when authors have the same role. Our results have implications for the design of systems supporting peer communication, e.g. peer-to-peer recommendation systems.