Eating disorder (ED) is a psychiatric condition that involves behaviors like binge and restrictive eating with severe health consequences, particularly prevalent among young women. While technology interventions exist, they typically focus on retrospective reflection or general management, missing the time window when an ED behavior is taking place. In this work, we conducted co-design sessions with 22 young women experiencing EDs to develop ideas for Just-in-Time (JIT) interventions, followed by interviews with five experts specialized in ED treatment. We found that eating plays varied roles in participants’ lives—from a means of gaining autonomy to automatic physiological responses—leading to design ideas including behavioral warnings, appetite management, food option redirection, psychological support systems, etc. By examining the characteristics of these designs with expert perspectives, we discuss what JIT support means for ED care and how to make it effective and sustainable. © 2026 Copyright held by the owner/author(s).

2025 CHI conference on Human Factors in Computing Systems, Yokohama, Japan, 26 April - 1 May 2025

Parents play essential roles in children’s play and learning with various media, often leading to positive and productive engagement outcomes for both parties. As such, an increasing number of HCI research has focused on understanding parent-child joint media engagement (JME) and designing new technologies to foster productive joint media experiences for children and parents. However, we currently lack a systematic view of this emerging field, which hinders the research and design of new joint media experiences and technologies for families. In this work, we conduct a scoping review of parent-child JME research within HCI (N = 89) and analyze the included papers from three lenses: publication features, methodological features, and JME features. Based on these findings, we identify gaps and opportunities in parent-child JME research and further expand the theoretical framing of JME by developing a framework that captures different JME dimensions.

Trustworthiness is an essential prerequisite for the real-world application of large language models. In this paper, we focus on the trustworthiness of language models with respect to retrieval augmentation. Despite being supported with external evidence, retrieval-augmented generation still suffers from hallucinations, one primary cause of which is the conflict between contextual and parametric knowledge. We deem that retrieval-augmented language models have the inherent capabilities of supplying response according to both contextual and parametric knowledge. Inspired by aligning language models with human preference, we take the first step towards aligning retrieval-augmented language models to a status where it responds relying merely on the external evidence and disregards the interference of parametric knowledge. Specifically, we propose a reinforcement learning based algorithm Trustworthy-Alignment, theoretically and experimentally demonstrating large language models' capability of reaching a trustworthy status without explicit supervision on how to respond. Our work highlights the potential of large language models on exploring its intrinsic abilities by its own and expands the application scenarios of alignment from fulfilling human preference to creating trustworthy agents.

The objective of this paper is to propose a hierarchical topic evolution model (HTEM) that can organize time‐varying topics in a hierarchy and discover their evolutions with multiple timescales. In the proposed HTEM, topics near the root of the hierarchy are more abstract and also evolve in the longer timescales than those near the leaves. To achieve this goal, the distance‐dependent Chinese restaurant process (ddCRP) is extended to a new nested process that is able to simultaneously model the dependencies among data and the relationship between clusters. The HTEM is proposed based on the new process for time‐stamped documents, in which the timestamp is utilized to measure the dependencies among documents. Moreover, an efficient Gibbs sampler is developed for the proposed HTEM. Our experimental results on two popular real‐world data sets verify that the proposed HTEM can capture coherent topics and discover their hierarchical evolutions. It also outperforms the baseline model in terms of likelihood on held‐out data.

Most of the existing approaches for image annotation generally demand exactly labeled training data, which are often difficult to obtain. In this letter we present a novel model that utilizes the rich surrounding text of images to perform image annotation. Our work makes two main contributions. First, by integrating text analysis, words that describe the salient objects in images are extracted. Second, a new probabilistic topic model is built to jointly model image features, extracted words and surrounding text. Our model is demonstrated to be flexible enough to handle multi-modal features and provide better performance than the state-of-the-art annotation methods.