TrustworthinessTransparencyEthicscs.HC
PeerPen: AI-Assisted Writing for Online Mental Health Peer Support
Jiwon Kim, Sherry Gong, Maya Ajit, Soorya Ram Shimgekar, Yunhao Yuan, Dong Whi Yoo, Eshwar Chandrasekharan, Koustuv Saha
September 14, 2026
Key Takeaways
-
AI writing assistance can lower barriers to peer support participation: The study found that PeerPen "reduced the burden of composing responses and increased confidence in offering support" among the 15 participants interviewed, suggesting that AI tools could help more people contribute meaningfully to online mental health communities.
-
Authenticity and personal connection are critical to peer support value: Participants expressed concerns that AI assistance could undermine the perceived authenticity of responses, which is central to why peer support works. The researchers note that "peer support derives much of its value from being perceived as personal," meaning AI tools must be designed carefully to preserve this essential quality.
-
Community-wide trust can be damaged even when AI isn't used: A key finding was that "such assistance could make authorship uncertain even for responses written without it, weakening trust across the community." This suggests that simply introducing AI writing tools into peer support spaces creates a trust problem for everyone, not just those using the technology.
-
Users want AI to assist, not replace, their voice: Participants "wanted AI to assist their writing without taking over authorship," indicating that people value maintaining control and ownership of their contributions. This points to the importance of designing AI as a collaborative tool rather than an autonomous agent.
-
Responsible AI design for peer support requires addressing authorship and transparency: The researchers emphasize the need for "design implications for AI writing assistance that scaffolds supportive communication, preserves authorship, and accounts for community-level trust," suggesting that technical solutions alone are insufficient without clear policies around disclosure and human agency.