Academics are squandering the potential of artificial intelligence to help with learning analytics, according to the architect of a programme that can determine at-risk students from the information they provide when they enrol.
A test of the AI model, developed at the Australian Catholic University (ACU), found it succeeded up to 17 times out of 20 in flagging international postgraduates at risk of disengaging from their studies.
Team leader Niusha Shafiabady, head of information technology at ACU’s Peter Faber Business School, said the students could be at risk of disengaging socially, intellectually or professionally as well as completely dropping out of their courses. She said universities’ attempts at early detection of such students tended to use traditional data analysis methods and to focus on single “dimensions” of engagement, with most existing tools focused on students’ grades and participation in classes. “It’s important to cover all the other aspects of engagement,” Shafiabady said. “[We need to] start using technology for good.”
ACU’s approach uses multiple machine learning techniques to gauge five types of engagement: personal, academic, intellectual, social and professional. Shafiabady said it relied on “very general information” about students including their age, country of origin, educational background, employment status, where they lived and how long they had been in Australia.
Universities could use information from the tool to “come up with pre-emptive actions” to help the students maintain motivation and curiosity, engage with their peers, improve their time management skills and seize internship opportunities, among other things.
A validation test of the model, outlined in the journal IEEE Transactions on Artificial Intelligence, found that it achieved accuracy rates of up to 85 per cent. “These findings provide valuable insights for both academics and policy makers, laying the groundwork for evidence-based strategies to improve student engagement,” the paper says.
A survey of more than 3,400 Australian university staff, published last year, found that the sector’s use of AI lacked sophistication. Universities were so focused on the risks presented by the technology that they tended to overlook its potential, particularly for administrative purposes.
Shafiabady said she hoped to develop her “AI brain” into a product that could be used by other universities.
The researchers trained the tool using information provided by 96 postgraduates at a city campus of an interstate university. It deemed most of the students “high risk” even though they appeared socially, intellectually and academically engaged.
The study found deficits in “personal” engagement – a reflection of individual motivation, time management and self-directed learning habits – and “professional” engagement through networking and work experience.
This suggested that many of the respondents – predominantly health master’s students from South Asia – were intellectually capable of completing their studies but not particularly interested in the subject matter. Shafiabady said a different sample would probably produce different results.
