Torrens University used generative AI to analyse and clean up the ‘look-and-feel’ 60,000 web pages of course content as part of a broader digital transformation.
Speaking to the iTnews Podcast, now-former director of transformation Eoghan Hogan said the novel use of generative AI enabled the university to uplift and standardise about five years’ worth of content.
While relatively young and digital-first, the university exists in a market that has undergone broad digital transformation in the past few years.
This led it to embark on a transformation of its own – “to get ahead of the game again”.
“We really want to have the best content in the best web experience for students,” Hogan said.
The university’s 21,000-odd students are evenly split between cohorts that study entirely online and those that attend face-to-face classes.
However, as Hogan notes, “all of our students have access to our digital learning assets.”
“It doesn’t matter where you’re studying or how you’re studying – all of our students have access to seven hours of on-demand learning assets per week, and they’re all delivered through the learning management system.”
Under the transformation, Torrens has set up an online student platform called MyLearn, that is powered by Instructure’s Canvas learning management system (LMS) and Microsoft Teams, with some “peripheral” systems also integrated to it.
MyLearn replaces a learning management system that was built on Blackboard.
MyLearn is the result of a significant body of planning work, which included a 16-week “discovery phase” involving students, alumni and competitive analysis that produced 1800 project requirements.
Not all of those requirements are reflected in MyLearn – some were allocated to different packages of transformation work – but those that stayed in reflected key demands from students for “superior UX/UI”, the prioritisation of video content, better mobile responsiveness and “persistent chat access” to course facilitators – all through a single platform.
“What they wanted is that kind of one single platform and then for us to really build their learning around that,” Hogan said.
Aside from needing to migrate the course content from one LMS to another, the university wanted to standardise the way it was formatted and presented.
As course content had been assembled and prepared by different subject matter experts over a number of years, the look-and-feel of it varied course-to-course, including the placement of objects on the page, and what digital assets a course offered for consumption.
Torrens had a team that had been experimenting with AI since 2022, and it suggested a novel way to examine the course content and formatting using generative AI.
Using Microsoft Azure OpenAI, it created a three-stage process for content uplift.
First, a bot scraped and identified the current content assets and placement. Then, prompts were used “to overlay a new UX/UI design, removing wasted space and reorganising the content”.
Finally, a human reviewed the work of the AI as a quality assurance measure to ensure the outcome was consistent.
Hogan said that initially, the university made use of the GPT-3 and GPT-3.5 large language models, but found they did not produce the desired accuracy in their output.
It was only when the university was invited to join an early access program to GPT-4 that the accuracy improved above 90 percent, and the conversion could be run at pace.
“We’ve now completed the first phase of the project,” Hogan said.
“We have been able to achieve 100 percent mobile responsiveness within our content.
“We’ve also been able to meet WCAG 2 0 [Web Content Accessibility Guidelines] from an inclusion and accessibility perspective, and we’ve also been able to improve the UX/UI consistency.”
Hogan said the university is trialling generative AI embedded within the LMS “to support their studies and their assessments”, with plans to roll that capability out to more broadly in future.