Deepfakes, Public Trust, and Online Discourse: Sentiment Analysis of Responses to the “Guru Beban Negara” Hoax Clarification on Social Media

Abstract

The rapid advancement of artificial intelligence technologies has introduced a new form of digital threat in the form of deepfakes, which have the potential to accelerate the spread of hoaxes and undermine public trust in official information. One notable and controversial example is the “Guru Beban Negara” hoax, which generated diverse and polarized reactions on social media platforms, despite the issuance of an official clarification by the authorities. This phenomenon suggests that institutional responses do not always succeed in mitigating negative public perceptions once misinformation has circulated widely. This study aims to examine public sentiment toward the official clarification by analyzing user comments on YouTube using the Naïve Bayes Classifier algorithm. Data were collected through web scraping techniques and subsequently processed through several preprocessing stages, including cleansing, normalization, tokenization, and stemming, to ensure data quality and analytical reliability. The findings reveal a strong predominance of negative sentiment among the 2,252 comments analyzed. The sentiment classification model achieved an accuracy rate of 73%, with the distribution indicating that most users expressed negative reactions, while only a small proportion conveyed positive sentiment. These results indicate that official clarifications have not been fully effective in alleviating negative public perceptions associated with deepfake-based hoaxes, particularly within the context of social media discourse.