YouTube as a helpful and dangerous information source for deliberate self-harming behaviours

Alhassan, Muhammad Abubakar and Pennington, Diane (2022) YouTube as a helpful and dangerous information source for deliberate self-harming behaviours. In: iConference 2022, 2022-02-28 - 2022-03-04, Virtual/Online.

[thumbnail of Alhassan-etal-iConference-2022-YouTube-as-a-helpful-and-dangerous-information-source-for-deliberate-self-harming-behaviours]
Preview
 Text. Filename: Alhassan_etal_iConference_2022_YouTube_as_a_helpful_and_dangerous_information_source_for_deliberate_self_harming_behaviours.pdf
Accepted Author Manuscript
License: Strathprints license 1.0
Download (1MB)| Preview

Abstract

Online social media platforms remain an excellent source of data for information scientists. Existing studies have found that people who self-harm find it easier to disclose information regarding their behaviour on social media as compared to in-person interactions. Due to the large and growing volume of user-generated content on YouTube, sources of videos presenting information concerning self-harm and discussions surrounding those videos could be hidden by other contents. By using a categorisation codebook and state-of-the-art topic and sentiment analysis techniques, the authors identified distinct groups of users who uploaded videos about self-harm on YouTube (n=107) and uncovered the topics and sentiments expressed in 27,520 comments. In addition to other sources, our investigations discovered that 56% of the people uploading the examined videos are non-professionals, in contrast to the group of professionals with only 11% of the videos in the sample. In grouping comments based on similar topics, we discovered that self-harming users, clean (recovered) users, at-risk audiences, and appreciative users responded to the examined videos. Viewers responded more positively to 'recovered from self-harm' and 'appreciative' responses, as opposed to 'at-risk' and 'self-harm' comments with a high negative sentiments. These features could be used to build a classifier, although more research is needed to investigate self-injurious information to better support digital interventions for effective prevention and recovery.

ORCID iDs

Alhassan, Muhammad Abubakar and Pennington, Diane ORCID logoORCID: https://orcid.org/0000-0003-1275-7054;
CORE (COnnecting REpositories)