Uncertainty and Support in Breast Health

Overview

Questions about breast health can arise after noticing a bodily change, arranging a consultation, undergoing tests, or preparing for treatment. This project examines how people bring those questions to an online community and what kinds of peer support they receive. I initiated the project to understand these experiences and develop research directions for supporting communication and care.

My role

I conceived and lead this project. I developed the research questions and study design, collected and prepared the data, conducted the LDA topic modeling, developed the codebook, carried out coding and statistical analyses, and wrote and revised the manuscript.

I am now extending the project through digital ethnography and outreach to clinicians and patients, exploring how human–AI interaction (HAI) interventions could be co-designed to support need expression, care communication, and support coordination.

Approach

I analyzed a corpus of 15,037 public Reddit threads. Latent Dirichlet allocation (LDA) identified ten discussion contexts, from discovering lumps and waiting for results to reading clinical reports. I constructed a stratified purposeful sample of 1,000 threads for content analysis of posts and replies.

The study examines expressed needs across four positions before treatment: appraisal, help-seeking, diagnostic investigation, and treatment preparation. Its two-stage framework distinguishes access to a human response from the presence of theoretically corresponding support categories in replies.

The current extension, Routing, Not Reading, explores how technology, potentially including AI, might help people articulate support needs and prepare questions for clinical conversations. Digital ethnography and outreach to clinicians and patients inform this developing co-design direction.

Findings

Earlier posts more often ask what a symptom or finding means. Questions about procedures and next steps, along with requests for emotional relief, become more prominent closer to diagnosis and treatment.

Receiving a reply does not necessarily mean receiving the kind of support a post calls for. Examining both response access and reply content makes these differences visible. In an exploratory analysis of 100 threads about reading clinical reports, 32 received a human reply and 13 contained informational support. These measures describe replies rather than whether a person’s needs were met or their health improved.

Manuscript

Title: From Symptom Appraisal to Treatment Preparation: Expressed Needs and Peer Support in an Online Breast Health Community

Status: Under review at the Journal of the American Medical Informatics Association (JAMIA).

Access: For a copy of the manuscript, please email ytpeng@unc.edu.

Related workshop paper

Title: Moving Beyond Symptom Search: How a Pre-Diagnostic Reddit Community Works Through Uncertainty

Status: Submitted to the CSCW 2026 workshop Women’s Health as Cooperative Work: Participatory AI Across Care, Community, and Policy.

Access: For a copy of the manuscript, please email ytpeng@unc.edu.