Current projects

My research focuses on designing and evaluating social computing systems that support real-world collaborative processes (e.g., self-help groups in online communities, co-designing) by harnessing machine intelligence.

Specifically, my research aims to design AI-infused systems that orchestrate both humans and AI to help people collaborate better.

(+ I like mixed-methods approach!)

↑ Hover over the diagram!

AI-generated design cards that translate scholarly articles for designers

Collaborators: Gary Hsieh, Lucy Lu Wang

A design card is a format of information that designers can use to conduct various design activities: understanding a phenomenon at hand, collaborative and playful ideation, co-design, formative evaluation of design concepts, and advocacy. (Fedosov et al., 2019)

Digging deeper into design cards led us to explore the following design question: how might we automatically transform the scholarly article to such actionable formats (i.e., design card) to support the design activities of designers - who are less familiar with research articles?

Chatbot for assisting collaboration of ad hoc teammates

Collaborators: Gary Hsieh, Soomin Kim, Ruoxi Shang, Joonhwan Lee

Many people gather online and form teams with strangers to collaborate on tasks. However, while intrateam trust and cohesion are critical for team performance, such characteristics take time to establish and are harder to build up through computer-mediated communication. Building on prior research that has shown that establishing common ground between members can help, we hypothesized that the use of a chatbot to support the familiarization of ad hoc teammates can help their collaboration. As such, we designed IntroBot, a chatbot that builds on an online discussion facilitator framework and leverages the social media data of users to assist their familiarization process. Through a between-subjects study (N=60), we found that participants who used IntroBot reported higher levels of trust, cohesion, and interaction quality, as well as generated more ideas in a collaborative brainstorming task. We discuss insights gained from our study, and present opportunities for the future of chatbot-assisted collaboration.

> CHI '23 paper

Exploring the roles of LLMs in supporting personal journaling processes

Collaborators: Taewan Kim, Young-Ho Kim, Hwajung Hong

Journaling can promote mental well-being by expressing personal thoughts and emotions without judgment of others. However, journaling can be challenging for people who find it hard to externalize internal states into words. Here, large language models (LLMs) may provide support in translating their ambiguous thoughts into writing and an opportunity to enrich people's vocabulary. Nevertheless, LLMs could disturb journaling by neglecting the personal context of users and even reduce users' initiative to write. In this study, we developed DiaryMate as a technology probe to explore the opportunities and challenges in using LLM in journaling to support reflection on one's experience. During the 10-day field study, participants perceived LLM as a tool to revisit their past experiences and as an emotional partner providing empathy. However, LLM sometimes disrupted participant's thinking and lowered autonomy. Based on the findings, we discuss the role of LLM in journaling and future design considerations.

AI-assisted supports in online mental health communities

Collaborators: Hwajung Hong, Jinwook Seo

Social support in online mental health communities (OMHCs) is an effective and accessible way of managing mental wellbeing. In this process, sharing emotional supports is considered crucial to the thriving social supports in OMHCs, yet often difficult for both seekers and providers. To support empathetic interactions, we design an AI-infused workflow that allows users to write emotional supporting messages to other users' posts based on the elicitation of the seeker's emotion and contextual keywords from writing. Based on a preliminary user study (N = 10), we identified that the system helped seekers to clarify emotion and describe text concretely while writing a post. Providers could also learn how to react empathetically to the post. Based on these results, we suggest design implications for our proposed system.

> CHI '22 Poster