Extract from ABC News
A number of universities have told a Senate inquiry it will be too difficult, if not impossible, to prevent students using AI to cheat assessments, and the institutions will have to change how they teach instead.
Key points:
- Universities have warned against banning AI technologies in academia
- Several say AI cheating in tests will be too difficult to stop, and it is more practical to change assessment methods
- The sector says the entire nature of teaching will have to change to ensure students continue to effectively learn
The tertiary sector is on the frontline of change coming from the rise in popularity of generative AI, technologies that can produce fresh content learned from massive databases of information.
Universities have widely reported experiences of students using AI to write essays or cheat assessments, with some returning to pen and paper testing to combat attempts to cheat.
In submissions to a Senate inquiry into the use of generative AI in education, a number now say it is not practical to consider attempting to ban the technologies from use in assessments.
Instead, some such as Monash University in Melbourne say the sector should "decriminalise" AI, and move away from banning it or attempting to detect its use.
"Beyond a desire to encourage responsible experimentation ... an important factor in taking this position is that detection of AI-generated content is unlikely to be feasible," its submission reads.
"Emerging evidence suggests that humans are not reliable in detecting AI-generated content.
"Equally, AI detection tools are non-transparent and unreliable in their testing and reporting of their own accuracy, and are likely to generate an intolerably high proportion of both false positives and false negatives."
Monash submitted that even if regulations were introduced to require AI tools to inject "watermarks" into their code to make AI detectable, other generative AI technologies could still be used to strip out those watermarks.
Instead, it and the nation's other largest universities under the Group of Eight (Go8) umbrella say the sector will have to change how it teaches and assesses students, using more oral or other supervised exams, practical assessments and the use of portfolios.
"Generative AI tools are rapidly evolving and will be part of our collective future – playing an important role in future workplaces and, most likely, our daily lives," tGo8 submitted.
"Entirely prohibiting the use of generative AI in higher education is therefore both impractical and undesirable."
The National Tertiary Education Union (NTEU) has also expressed skepticism that universities would be able to completely manage AI misconduct — not only in assessment, but also in research.
"There is a real risk that AI applications will be considerably ahead of current research integrity processes that would detect problems or irregularities," the NTEU submitted.
"It may be that problematic research is not detected for some time, if at all, by which time there could be widespread ramifications."
New system needed to ensure students are actually learning
Universities have also given evidence that they have begun using generative AI across every aspect of what they do.
Monash University described how it has tested using AI "personalised course advisers" to help students navigate their degrees and classes, AI-powered mock job interviews for real positions and simulated customers or clients for learning.
"For example, in health education, the tool provides realistic ‘patients’ with detailed medical histories, personas, and varied willingness to share embarrassing medical details with learners who must put the work in to develop rapport with the patients to obtain relevant information in a realistic virtual clinical environment," it said.
As generative AI begins to be picked up at every part of the learning cycle, the tertiary education union warned that, over time, there was a risk universities may reach a point "where they can no longer assure the required learning has occurred in what they claim to be teaching".
"Teaching staff will need to continuously develop new methods of assessment that assess students at a level beyond the levels of AIs," it said.
The Queensland University of Technology said unless the nature of learning changed, there was a risk AI could promote "laziness and lack of independent thought".
"The possibility of a situation [arises] in which activities are created by the educators using AI, the learners use AI to create their responses, and the educators use AI to mark/grade and even give feedback," Queensland University of Technology wrote.
"At its most extreme, such a scenario suggests the question of who, if anyone, has learnt anything? And what was the purpose of the assessment?"
No comments:
Post a Comment