Identification of AI Generated Texts Using an Ensemble Method

Since the rise in popularity of artificial intelligence, it has found its way into many aspects of everyday life which we once thought to be quite impossible. Today, AI tools are commonly used to generate various types of text, including academic papers and business reports. Although AI has shown remarkable potential and capability in producing text, it is increasingly important to set apart between content written by humans and that generated by AI, especially to protect against misuse and fraud. In this study, we explored an ensemble learning approach using three different BERT-based embeddings, achieving results that scored almost perfectly on the evaluation metrics. The code used in our experiments will be publicly shared to allow others to verify and replicate our findings.