MHMisinfo - Video-based Mental Health Misinformation Dataset

MHMisinfo-Gold and MHMisinfo-Large datasets, as described in the paper "Supporters and Skeptics: LLM-based Analysis of Engagement with Mental Health (Mis)Information Content on Video-sharing Platforms" (forthcoming at ICWSM 2025). Videos and comments for each dataset are seperately stored in different .csv files

Dataset schema, videos

Column Name
Description

video_id
ID of the Video, as assigned by their respective platforms

video_title
The title of the video

video_description
The description of the video, given by the video creators

audio_transcript
Text transcription of the video's audio track, as generated by Whisper speech-to-text model

video_view_count
View count of the video, at the time of data collection

video_like_count
Like count of the video, at the time of data collection

video_comment_count
Comment count of the video, at the time of data collection

label_ioi
"Information of Interventions" label of video, annotated by experts. 1 = High-quality information on interventions, -1 = Low-quality information on interventions

label_ebt
"Evidence-based Treatment" label of video, annotated by experts. 1 = Encourages evidence-based treatment, -1 = Discourages evidence-based treatment

label_aoc
"Alignment of Consensus" label of video, annotated by experts, 1 = High Alignment with Consensus, -1 = Low Alignment with Consensus

label
Overall mental health misinformation label of the video. 0 = non-MHMisinfo videos, and -1 = MHMisinfo videos

platform
Platform of the video

Dataset schema, comments

Column Name
Description

text
The raw text of the comment

commenter_channel_display_name

The display name of the user who posted the comment.

comment_publish_date

The time when the comment was orignally published, .

video_id

ID of the Video associated by the platform, as assigned by their respective platforms

platform

Platform of the video associated with the comment

label

Overall mental health misinformation label of the video associated with the comment. 0 = non-MHMisinfo videos, and -1 = MHMisinfo videos