This study investigated whether ChatGPT emotional support was viewed as supportive, would encourage people to seek professional help, and whether Hikikomori, extreme social withdrawal, and internet addiction moderated this effect. We evaluated participants' Hikikomori levels, internet addiction levels, and pre-experiment desire to seek help. Participants then read a brief online conversation of an individual sharing their worries with ChatGPT. The chat content was identical across conditions, except participants were randomly told that the responses were from ChatGPT or a hotline professional mental health counselor who provided temporary help in the face of crisis. Lastly, participants rated perceived message quality and emotional support, and post-experiment desire to seek help. We hypothesized that participants would rate ChatGPT's messages with lower message quality and emotional support than human's messages (H1); Hikikomori and internet addiction would moderate perceived emotional support and message quality (H2); Individuals would desire to seek help more after viewing ChatGPT's messages across conditions (H3). 417 participants, 18 - 28 years old, living in the U.S. were recruited from CloudResearch Connect in July 2025. We found that both groups' desire to seek help increased post-experiment, and rated message quality and emotional supportiveness similarly, regardless of Hikikomori levels. Internet addiction may moderate the effect of message source on emotional support. The presentation will discuss on the implications, limitations, and future directions of the study.
Primary Speaker
Mai Tran
Faculty Sponsors
Catherine Walker
Presentation Type
Faculty Department/Program
Faculty Division
Do You Approve this Abstract?
Approved
Time Slot
Room
Topic
Session
Moderator
Catherine Walker