Analyzing Global Attitudes Towards ChatGPT via Ensemble Learning on X (Twitter)

Abstract

This research investigates global public attitudes towards ChatGPT by analyzing opinions on X (Twitter) to better understand societal perceptions of generative artificial intelligence (AI) applications. As conversational AI systems become increasingly integrated into daily life, evaluating public sentiment is crucial for informing responsible AI development and policymaking. Unlike many prior studies that adopt a binary (positive-negative) sentiment framework, this research presents a three-class classification scheme-positive, neutral, and negative framework, enabling more comprehensive evaluation of public attitudes using X (Twitter) data. To achieve this, tweets referencing ChatGPT were collected and categorized into positive, neutral, and negative opinions. Several algorithms, including Naïve Bayes, Support Vector Machines (SVMs), Random Forest, and an Ensemble Learning model, were employed to classify sentiments. The Ensemble model demonstrated superior performance, achieving an accuracy of 86%, followed by SVM (84%), Random Forest (79%), and Naïve Bayes (66%). Notably, the Ensemble approach improved the classification of neutral sentiments, increasing recall from 73% (SVM) to 76%, underscoring its robustness in handling ambiguous or mixed opinions. These findings highlight the advantages of Ensemble Learning techniques in social media sentiment analysis and provide valuable insights for AI developers and policymakers seeking to understand and address public perspectives on emerging AI technologies such as ChatGPT.