A SENTIMENT ANALYSIS OF YOUTUBE VIDEOS FROM DONOR-CONCEIVED PEOPLE, UTILIZING ARTIFICIAL INTELLIGENCE (CHATGPT)

To assess the sentiments of donor-conceived people (DCP) towards their life experiences in videos uploaded on YouTube, using chatGPT, an Artificial Intelligence natural language processing tool. A cross-sectional analysis of YouTube videos identified via the search term “donor conceived” was conducted. While using a new YouTube profile on a private browser, two independent reviewers performed the search on March 15, 2023. The initial 200 videos sorted by relevance were reviewed. Only English-speaking videos and those describing DCP’s life experiences via autobiographical, interview, or documentary format were included. Excluded videos were those told from perspectives other than those of DCP, YouTube "Shorts", and those without English-language transcripts. Disagreements in video inclusion and exclusion were resolved using a standard protocol. Video characteristics, including year of upload, video length, number of views, number of comments and country of origin were obtained. Video transcripts were input into ChatGPT (https://chat.openai.com) using the following query: “Given this text, what is the sentiment conveyed? Is it positive or negative? Text: {transcript}”. Sentiment results were categorized as neutral, positive, negative, or mixed. ChatGPT limits queries to 3,000 words; therefore transcripts containing greater than 3,000 words were separated into multiple queries. The most commonly occurring sentiment was then selected as the overall sentiment of the transcript. Chi-square goodness-of-fit and independence tests were performed to assess observed proportions of sentiments compared to expected distribution. 144 YouTube videos were identified. Video length ranged from 30 seconds to 113.52 minutes. Number of views ranged from 3 to 6,747,090 views. 22.9% of videos were determined to have a neutral sentiment, 27.8% positive, 25.7% negative, and 23.7% mixed (p=0.84). The median number of views was 392.5. Of videos with less than 392.5 views, 25% had positive sentiment and 32% had negative sentiment, and of those with more than 392.5 views, 31% were positive and 19% were negative (p=0.13). The median video length was 10.24 minutes. Of videos less than 10.24 minutes in length, 33% were positive in sentiment and 24% were negative, and with videos greater than 10.24 minutes in length, values were 22% and 26%, respectively (p=0.26). The sentiments of DCP YouTube videos are generally evenly distributed between neutral, positive, negative, and mixed. Video views and length of video were not associated with a significant difference in sentiment distribution.