AI In Education | Insights from an Academic Partner

August 2025

Artificial intelligence and large language models like ChatGPT are rapidly evolving in both capability and application. At Monash University, we have been proactively exploring how to incorporate these tools into our curriculum—helping students understand their strengths, limitations, and potential pitfalls.

Within the Monash Institute of Transport Studies, several educators are already integrating AI tools to enhance student learning.

When A/Prof Nan Zheng interviewed students about using ChatGPT or similar Large Language Models (LLMs), students said they were “helpful but not that helpful." They recognised that these tools often provided unreliable results, for example, "for design-related calculations, the LLM was like a confident friend who didn't really understand the knowledge". Thus, most students treated LLMs as a “one-time solution generator” for seeking quick answers but rarely engaged further.

A/Prof Alexa Delbosc has been helping students overcome this “not helpful” performance by teaching students how to improve the quality of the prompts they use to ask questions. She is using the case study of preparing risk assessments for traffic engineering fieldwork, a crucial skill required by Engineers Australia. In addition to traditional teaching guidance on how to assess risk, in 2025 Dr Delbosc invited students to use ChatGPT to generate a first draft. The students quickly learned that writing a detailed prompt was a crucial first step to getting a meaningful answer – although not all answers were meaningful. Although only some students went on to use these drafts in their final risk assessments, the overall quality of work this year was higher than previous years.

I had a colleague check the risk assessments from three years’ worth of classes, without knowing which ones used ChatGPT,” says A/Prof Delbosc. The risk assessments from 2025 scored higher across the board, especially on identifying potential hazards that previous years’ students may not have considered. However, the findings were not all positive, as some students included hazards that were not relevant to the required fieldwork.

A/Prof Zheng is leading a team of PhD scholars – Linxin Hua and Lirui Guo – who are going a step further to create a custom “learning assistant” LLM trained on road engineering learning materials. Initial testing has shown that customising the learning tool can make a remarkable difference. According to the students, the answers generated by the custom tool are more relevant and “grounded in actual course materials and the required industry standards". Furthermore, students were more willing to engage with the system, to ask follow-up questions or to rephrase challenging questions. Currently, Linxin and Lirui are conducting comprehensive large-scale testing on the learning assistant in real-world educational settings. Beyond technical validation, the team also aims to explore how AI-enhanced learning tools can reshape the education pedagogy for engineering in general.

Monash academics continue to explore ways that AI learning tools can enhance learning rather than replace human judgement. A/Prof Delbosc asks her students, “At the end of the day, if ChatGPT could do this task without you, why would anyone hire you to be a traffic engineer?”