This chapter presents an innovative approach to analysing survey data on the ethical use of AI tools, such as ChatGPT, in academia. Using a mix of closed- and open-ended survey questions, we collected diverse perspectives from graduate students from different cultural backgrounds. These students were enrolled in graduate management programmes on two campuses of a top international university. Closed-ended questions are efficient for data collection and analysis but may limit exploration, while open-ended questions provide deeper insights but require more resources for analysis. To analyse the open-ended responses more efficiently, we used Structural Topic Modelling (STM), a machine learning technique that identifies themes in open-ended responses and assesses the influence of covariates, such as gender and culture, on those themes. We also used ChatGPT to classify cultural clusters according to a taxonomy from existing research. The main contribution of this chapter is methodological. It provides detailed guidance on the methods used to help other researchers replicate the approach. In addition, this chapter also offers valuable insights into the ethical considerations of using AI tools in education.

错误:搜索内容不能为空,请输入英文关键词
错误:关键词超出字数限制,请精简
高级检索

A Detailed Guide to Structural Topic Modelling in Educational Research: Insights from a Study on the Ethical Use of AI in Academia

  • Michael Weiss,
  • Sabiha Mumtaz,
  • Jamie J. Carmichael

摘要

This chapter presents an innovative approach to analysing survey data on the ethical use of AI tools, such as ChatGPT, in academia. Using a mix of closed- and open-ended survey questions, we collected diverse perspectives from graduate students from different cultural backgrounds. These students were enrolled in graduate management programmes on two campuses of a top international university. Closed-ended questions are efficient for data collection and analysis but may limit exploration, while open-ended questions provide deeper insights but require more resources for analysis. To analyse the open-ended responses more efficiently, we used Structural Topic Modelling (STM), a machine learning technique that identifies themes in open-ended responses and assesses the influence of covariates, such as gender and culture, on those themes. We also used ChatGPT to classify cultural clusters according to a taxonomy from existing research. The main contribution of this chapter is methodological. It provides detailed guidance on the methods used to help other researchers replicate the approach. In addition, this chapter also offers valuable insights into the ethical considerations of using AI tools in education.