Still need help?
Get in touch with the LX.lab team by logging a ticket via ServiceConnect. We'll be in touch shortly.
Log a ticketWant to provide feedback on this resource? Please log in first via the top nav menu.
A lecturer at the School of Sport, Exercise and Rehabilitation shows how highlighting the weaknesses of AI will get students thinking.
Academic in focus
Michael Rennie is a lecturer at the School of Sport, Exercise and Rehabilitation. Michael graduated from UTS in 2011 and completed doctoral studies in team sport analysis while working with the Sydney Swans Football Club.
Mike is the subject coordinator of 96301 High Performance Science Autumn 2023, which examines the role that science plays in enhancing human performance in high-performance settings.
The challenge
ChatGPT is able to provide seemingly decent answers to assessments, but lacks scientific accuracy. How can this be harnassed to benefit students?
Objective
To use AI’s lack of insightful analysis as a backdrop for students to provide evidence of understanding. This subject teaches students how to assess the quality of information, which includes AI, to support decision making when providing new interventions to high performance athletes.
The task
Students will learn to evaluate the opinions of expert sports science practitioners, as well as how to develop and provide a graded recommendation on the suitability of new interventions. Students will then use AI to generate responses to compare and contrast with the expert opinions.
Students will be provided with resources and scaffolding activities as a module in Canvas prior to them attempting the activity above.
AI results
ChatGPT was tested to:
- write an exercise program to build muscle for a college-aged student. Within 30 seconds there was a comprehensive program generated.
- compare and contrast two views on the theory of fatigue. It gave a pretty good comparison.
- compare research on using metrics for predicting injury. There are two schools of thought about injury and it was really good at summarising the arguments, but unfortunately it attributed the wrong authors to the wrong side of the argument.
What next?
- Students need to be able to understand the limitations of AI. There are certain areas where the responses were well-generated by AI and others where there were limitations. In one test, the reference list that was generated by AI was quite generic.
- You can ask ChatGPT to write an assignment and it will come up with a reasonable answer, but it won’t be at a scientific level. ChatGPT doesn’t access scientific databases for now.