Key points are not available for this paper at this time.
This study aimed to examine consumer perceptions of artificial intelligence (AI) by analyzing YouTube comments using topic modeling. A natural language processing model called BERTopic was employed to identify latent topics in a total of 93,729 English-language comments. These comments were posted on 20 popular AI-related videos, each of which had over 1 million views, between January and December 2021. The analysis of the comment data yielded five major themes, each with its corresponding percentage representation. The first theme, accounting for 27.2% of the comments, revolved around the fear of AI superiority, reflecting concerns about AI surpassing human control. The second theme, comprising 13.4% of the comments, focused on anthropomorphism and the uncanny valley effect, indicating mixed emotions towards human-like robots. The third theme, representing 31.7% of the comments, highlighted AI bias and hallucinations, shedding light on the impact of AI inaccuracies on trust. The fourth theme, accounting for 17.2% of the comments, addressed ethical and moral concerns, particularly concerning AI development and oversight. Lastly, the fifth theme, encompassing 10.5% of the comments, explored AI trustworthiness, emphasizing the reliability of AI in decision-making processes. Based on the findings, this study challenges previous research that primarily reported positive AI experiences. It quantitatively identifies complex negative emotions towards AI technologies across diverse user demographics. Furthermore, it expands our understanding of how anthropomorphism and AI "hallucinations" influence trust, which are critical issues as AI becomes increasingly integrated into society. The application of BERTopic enabled nuanced modeling of online discourse, bridging gaps in our knowledge regarding consumer perceptions of emerging technologies on a large scale. In terms of novelty, this study is the first to utilize BERTopic modeling to uncover consumer experiences with AI by analyzing social media comments on a large scale. The insights obtained from this approach offer compelling new perspectives compared to previous research in this field.
Sohaib Mohanna (Wed,) studied this question.