Growing concern about increased AI use in universities

[Hanno Labuschagne]

Growing concern about increased AI use in universities

Concerns about the growing use of artificial intelligence (AI) tools by university students are mounting, and universities in South Africa may have to transition from written assignments as a result.

According to AI engineer Akhil Boddu, traditional written assessments may no longer accurately reflect whether students actually understand their coursework.

 

[garyc]

This is probably not an insurmountable problem for undergraduates, where they will eventually have to sit in a supervised exam.

There may be a bigger problem with postgraduate thesis generation since this is largely unsupervised. Everyone is now using tools like NotebookLM, even the academic staff, to carry out more insightful literature surveys. A person may even find themselves at a disadvantage if they do not use it. The problem arises if they then start using generative AI to develop significant parts of their thesis. I experimented with doing this about a year ago and this can generate very good results. Using this tactic is also becoming harder to detect. The AI tries to mimic human academic writing and it is getting better all the time.

 

[Kosmik]

According to AI engineer Akhil Boddu, traditional written assessments may no longer accurately reflect whether students actually understand their coursework.

“The purpose of assignments is to validate knowledge. Do you actually understand a concept? The assignment validates that,” he told Cape Talk.

Boddu said universities in South Africa may need to transition to new assessment methods that focus more on demonstrating understanding rather than regurgitating facts.

“They have to demonstrate their understanding,” he said.

He proposed changing assessments to include oral interviews after assignments and requiring students to clearly indicate how they’ve used AI in their assignments.

TBH, this been a tertiary issue for years, AI has just made it much more widespread. There should be a panel or presentation component to validate knowledge. IT had things like brain dumps etc for decades.

I've always admired open book tests because those are a true test of application of knowledge, not whether you memorize some random formulae or piece of information.

Blocking AI usage is stupid in this day and age, like blocking library or research access etc. Make them prove their knowledge and understanding, not parrot it.

 

[Smiggles]