CONTENT WARNING: This paper discusses image-based sexual abuse. Technological advances in Artificial Intelligence (AI) have made it easy to generate and distribute Non-Consensual Synthetic Intimate Imagery (NCSII); images and videos that depict people’s voices, faces, or bodies in intimate or sexually explicit scenarios. The creation and distribution of NCSII is a form of image-based sexual abuse that primarily affects marginalized people. We present Photo BOO-th, an interactive installation that invites attendees to encounter NCSII of themselves, resulting in a creepy, visceral, and affective experience. We designed Photo BOO-th to critique, raise awareness, and foster societal discussions around consent, image-based sexual abuse, and the role of technology in enabling harm. Following a Research through Design approach, we unpack design events that reveal some of the tensions and considerations in designing Photo BOO-th. We conclude with a discussion around designing creepy interactions in sensitive contexts, designing with and against the uncertainties in generative AI, and a reflection on vulnerability when designing creepy interactions.

We invite a visual exploration and critique of imagery that can be (re)produced through open-source models available in Civitai — a repository where creators share and monetize custom text-to-image models. The proliferation of these models contributes to the creation and distribution of Non-Consensual Synthetic Intimate Imagery (NCSII), a digitally-mediated form of image-based sexual abuse that primarily affects women and girls, commonly referred to as “deepfake porn.” We examine the practices surrounding the creation and sharing of custom open-source models designed to produce photorealistic content. Through a qualitative content analysis of 510 models and 3800 images, we illustrate the imaginaries about sex, bodies, and gender expressions that models embody and perpetuate, as well as how model creators understand their practices in relation to potential harms and misuse. We discuss pathways for feminist intervention and the care practices we embedded into our research process, and the design of this pictorial.

How do product teams evaluate LLM-powered products? As organizations integrate large language models (LLMs) into digital products, their unpredictable nature makes traditional evaluation approaches inadequate, yet little is known about how practitioners navigate this challenge. Through interviews with nineteen practitioners across diverse sectors, we identify ten evaluation practices spanning informal ‘vibe checks’ to organizational meta-work. Beyond confirming four documented challenges, we introduce a novel fifth we call the results-actionability gap, in which practitioners gather evaluation data but cannot translate findings into concrete improvements. Drawing on patterns from successful teams, we contribute strategies to bridge this gap, supporting practitioners’ formalization journey from ad-hoc interpretive practices (e.g., vibe checks) toward systematic evaluation. Our analysis suggests these interpretive practices are necessary adaptations to LLM characteristics rather than methodological failures. For HCI researchers, this presents a research opportunity to support practitioners in systematizing emerging practices rather than developing new evaluation frameworks.

This paper critically examines the machine learning (ML) modeling of humans in three case studies of well-being technologies.Through a critical technical approach, it examines how these apps were experienced in daily life (technology in use) to surface breakdowns and to identify the assumptions about the "human" body entrenched in the ML models (technology design).To address these issues, this paper applies agential realism to decenter foundational assumptions, such as body regularity and health/illness binaries, and speculates more inclusive design and ML modeling paths that acknowledge irregularity, human-system entanglements, and uncertain transitions.This work is among the first to explore the implications of decentering theories in computational modeling of human bodies and well-being, offering insights for more inclusive technologies and speculations toward posthuman-centered ML modeling.