SQA exploring use of AI to analyse exam marker feedback

The Scottish Qualifications Authority is exploring whether artificial intelligence could play a role in helping to identify and analyse key themes in feedback from exam markers.

The hope is that careful use of AI could lead to any issues being addressed earlier in the “intensive” period after exams.

The move comes after the fallout over the drop in the Higher history pass rate in 2024.

Higher history paper two, on Scottish history, was the focus of much of the controversy in 2024 and will be marked via digital images this year - as opposed to the traditional approach of marking paper scripts.

The SQA said that when marking online there is “an ongoing feedback loop established to allow any issues arising from understanding or application of the marks scheme to be addressed in real time”.

SQA: Quality assurance ‘ongoing’

It also said that “quality assurance is ongoing” and “there is an inbuilt escalation process for markers and team leaders”.

However, Higher history paper one, which focuses on British, European and world history, will not be marked “from image” until 2026.

In 2024, 72 per cent of externally assessed elements of national qualifications were marked from image. In 2025 that figure is anticipated to increase to 75 per cent, and in 2026 to 82 per cent.

• Background: Higher history review finds marking standard did not change
• Analysis: Higher history controversy underpins case for independent exam watchdog
• News: History teachers give ‘damning’ evidence in row over SQA grading

Last year the Higher history pass rate fell by just over 13 percentage points on 2023, leading to complaints from parents and teachers that the marking standard had changed.

An SQA review in November found that the marking standard in 2024 did not change and that marking and grading processes worked as intended. It put the sharp drop in Higher history passes down to “the poor standard of responses” from students, especially on the Scottish history paper.

The review identified “areas for continuous improvement”, where action could “strengthen” the assessment and awarding process. This included reviewing the way marker feedback is considered and acted upon.

It stated: “This should help to mitigate concerns that issues markers have experienced during marking have not been addressed at later stages of the quality-assurance process before final awards are made.”

Although the review found that feedback from markers “overwhelmingly focused on the poor standard of responses provided by learners”, it also listed feedback from some markers who felt the standard for paper two, on Scottish history, had changed in 2024.

One marker said that the marking standard “was much higher than previous years”; another said it “appeared to be much stricter…for certain topics”. A third commented: “The marking was so harsh this year, particularly in the ‘explain’ question. It felt like the goalposts were moved after the exam.”

Risks and rewards of AI

Now the SQA has shared the action it has taken in response to the review’s recommendations with the Scottish Parliament’s Education, Children and Young People Committee. The 7 March update was also shared with Tes Scotland following a freedom-of-information request.

In terms of marker feedback, and ensuring that this is “considered and used appropriately” in the future, the SQA stressed that any changes will have to “minimise any additional demands” on the core examining team and the principal assessor, given that the post-exam period is “already very intensive and pressurised”.

It added: “One key area for exploration will be the way in which feedback is gathered from markers to allow it to be analysed and key themes identified more readily. As part of this the working group will be asked to explore how AI may be able to contribute to this, and any risks associated with its use in this context.”

An update in February advised that initial discussions were “underway with SQA staff, who are also looking at AI for collating and analysing data for research and informative purposes”.

It added: “This will help to inform longer-term changes to analysing feedback from marker reports.”

The February update also acknowledged that “there are risks involved in using AI where there is confidential information involved”.

For the latest in Scottish education delivered directly to your inbox, sign up for Tes’ The Week in Scotland newsletter