AI for AOs: What Ofqual Guidance Says About GPT Plugins
Find out if Integrity Advocate is right for you
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Get a Free Demo

AI for AOs: What Ofqual Guidance Says About GPT Plugins

AI for AOs: What Ofqual Guidance Says About GPT Plugins

Scour the current Ofqual Handbook (we did) and you won’t find AI or ChatGPT mentioned. A more recent publication, a 2023 research paper on Remote invigilation within vocational and technical qualifications, only states in passing that “this research took place at a time when generative artificial intelligence tools such as ChatGPT were not widely available and so the challenges such tools present have not been explored.”

What the Handbook does require is for Awarding Organisations (AOs) to ensure the integrity of their processes. And yet, as we pointed out in a recent blog post, AI study aids are an increasingly prominent threat to this integrity, especially when it comes to remote exams and training. 

Understanding the Threat

GPT-powered ‘study aid,’ ‘homework helper’ and ‘learning companion’ Chrome plugins automatically generate correct answers in online exams. These plugins can be made undetectable on-screen, and some even go so far as to obscure their code on the backend. 

With a plugin activated, the experience for learners is simple: open their exam and… all the questions are answered automatically, without having to click, read or leave the page.

Effectively, these plugins allow anyone to have an exam completed for them, without an outside observer knowing the difference. For obvious reasons, this has made it extremely challenging for Awarding Organisations to maintain the integrity of any qualifications that are delivered online.  

No Easy Answers

So where are AOs to look for guidance when it comes to managing these threats? To start, it’s important to remember that Ofqual guidelines are just that: guidelines. They provide advice and frameworks for thinking about remote testing, but they don’t offer up answers. And, in a way, when it comes to preventing unethical AI use, there are no easy, one-size-fits-all answers. 

If you’re going to invest in online testing, the experience needs to be easy for students. But turning to invasive invigilation systems to prevent the use of AI becomes very complicated, very quickly. Support requests go up. Costs balloon. Complaints multiply. What’s an AO to do? 

Look to the Guidelines

While the Ofqual Handbook doesn’t mention AI, several conditions are broadly applicable to testing integrity as a whole, and as a result have implications for AI use. The most relevant section here is Condition 8: Completion of the assessment under the required conditions:

An awarding organisation must take all reasonable steps to ensure that, in relation to qualifications which it makes available… evidence generated by a Learner in an assessment is generated by that Learner.

The takeaway is simple in theory, but complex in execution: not only do participants need to have their IDs verified prior to taking the exam, they must also be monitored for things like cell phone use, on-device communication tools (video conference, chat, etc) and AI-enabled browser plugins that can automatically answer questions without their input.

Monitoring for AI

The most common way for AOs to monitor for AI usage is to track and review test completion times. If the average learner completes an exam in 65 minutes and receives a score of 85%, any significant deviations from this average — say, learners who complete the exam in 10 minutes and get 100% — could indicate the use of an AI plugin.

Unfortunately, test-takers are catching on to this. We’ve seen numerous Reddit forums where students share tips on disguising AI usage, including delaying submitting their exams and deliberately answering a few questions wrong to throw invigilators off the scent. 

As a result, AOs need to change their tactics, too. Rather than being reactive (looking for evidence of AI use after the fact) be proactive and stop it before the exams begin. 

Preventing Plugin Use

In the world of invisible AI plugins, the best way to maintain test integrity in accordance with Ofqual guidelines is to prevent access to assessment content if the learner has an AI plugin active on their device. 

There are several tools on the market that claim to do this; our product, ExposeAI, is the only one that does not require installation by the learner (as we like to say, you can’t fight a plugin with a plugin). ExposeAI continuously monitors for AI plugin use; if one is found, ExposeAI can automatically block access to the exam until it is deactivated. 

The Future of AI and Ofqual

An update from Ofqual is apparently on the way which will clarify expectations and guidance around AI monitoring. Until then, however, Awarding Organisations must remain proactive; you can’t know what the future will hold, but you can look to the guidance that’s available now and anticipate how to best position yourself.

Because either way, neither AI nor online testing is going away anytime soon. The more you can stay ahead of the curve and maintain the integrity of your programming, the less stress you — and your learners — will experience along the way.