With the emergence of large language models (LLM) and Artificial Intelligence (AI) assistants like ChatGPT, accompanying tremendous potentials are critical challenges. Indeed, these assistant systems can provide quality information with conveniences. However, the generated contents are highly problematic being seemingly indistinguishable from that of human. The implication of this issue is severe in science, education, and domains that value original contents. With such motivation, this project addresses the task of identifying ChatGPT-synthesized texts with a focus on education, specifically, in short-answer questions. The goal of the project is to develop an AI technology that identifies synthesized texts by comparing such contents to examples known to be from AI for the same questions.