Our updated position statement on artificial intelligence

The emergence of generative AI models in autumn 2022, to which access has continued to broaden, has inspired significant interest in their role in learning and qualifications. There has also been consideration of whether other forms of AI could play a role in the qualification system. The capabilities of AI models continue to grow, and mean that they can be used for different purposes, such as generating text and visuals in response to user criteria, or processing large amounts of data to inform predictions and language translations. 

While the applications of AI are wide ranging, the outputs created by AI models are not always reliable. The UK government’s 2023 white paper on AI regulation suggested that AI models could be defined as operating through a combination of adaptivity and autonomy, with these characteristics meaning it can be difficult ‘to explain, predict or control the outputs of an AI system’. This means that care needs to be taken before deciding whether AI models should be used in qualifications and assessments.  

As we consider the role of AI in qualifications, we will consider how this supports the achievement of our principle aims. We will also have a duty to carry out sustainable development from 1 April 2025, as we become a public body subject to the Well-being of Future Generations (Wales) Act 2015. Our wellbeing goals, through to 2035, are to: 

• shape qualifications that prepare and support learners in life, learning and work, and promote equality of opportunity
• develop an agile qualification system for Wales that responds to socioeconomic changes, meeting learners’ needs whilst protecting the value of the qualifications they receive
• operate as a forward-thinking, high-performing and inclusive organisation that delivers for the people of Wales

The emergence of AI models and their ongoing advances bring both opportunities and challenges to the qualification system. We are positive about the potential benefits that AI models could offer, and recognise that they could play a role in the achievement of our well-being goals. We also see that these models present challenges to how learners are assessed fairly, and we want to ensure that the integrity of qualifications is protected.

In considering the impact of the wide availability of AI models, we have identified four broad areas which present potential opportunities and challenges:

• the knowledge, skills and experiences contained within qualifications, and how these are kept up to date
• the use of AI in learning and teaching 
• where and how candidates take their assessments
• assessment and awarding processes 

2. The delivery and assessment of existing qualifications

We have taken steps to understand how the availability of AI models could impact the delivery and assessment of existing qualifications by awarding bodies and learning providers. We have liaised with Welsh Government, and we meet regularly with AI leads from Ofqual in England and CCEA Regulation in Northern Ireland. This allows us to share learning and developments which are relevant to qualifications offered across different areas of the UK.  We also collaborate with fellow qualifications regulators on our Annual Statement of Compliance activity, allowing us to better understand how awarding bodies are managing potential risks associated with AI models.

We have supported the work of the Joint Council for Qualifications (JCQ), which has developed and refined guidance for teachers and assessors on the appropriate use of AI in non-examination assessments. Unlike examinations, which are usually taken under highly controlled conditions which do not permit the use of the internet, non-examination assessments typically offer greater levels of flexibility. This means that further planning and consideration may be required to ensure the fairness of non-examination assessments. JCQ’s guidance supports centres to do this, including by covering: 

• examples of AI misuse
• steps for preventing and identifying misuse
• marking assessments where AI has been used

We have engaged with learning providers, including through the Jisc Wales AI Working Group, to develop our understanding of how centres and learners are using AI in teaching and learning. We have also seen examples of the AI policies and training that centres are developing and implementing, and useful guidance for learners and for staff in further education settings.

We have commissioned a research study, part of which explores stakeholders’ views on the risks and opportunities associated with AI and qualifications. The study asks whether new qualifications on the subject of AI are needed, about the use of AI in assessment processes, and about issues such as academic integrity. Additionally, we are conducting an internal study about the implications of AI use in delivering assessments. 

Our engagement with awarding bodies has indicated that many are giving active consideration to how AI technologies can be used in their delivery of qualifications. We are also aware that Ofqual has issued clarifications on specific uses of AI against its conditions, such as confirming that using AI models as a sole marker of learners’ assessment evidence would not be compliant due to the potential for inaccuracy. Recognised awarding bodies in Wales have a legal requirement to comply with our Standard Conditions of Recognition, and we consider AI technologies to be a relevant aspect of compliance against these conditions. While we encourage awarding bodies to innovate when they can, some specific applications of AI technologies are currently likely to lead to non-compliance with our conditions, such as AI models being used:

• to develop assessments without human review
• as the sole method of remotely invigilating assessments
• to fulfil the role of an Assessor (defined on page 68 of our Standard Conditions of Recognition)

3. AI technologies in qualifications in the future

We are engaging widely to understand the opportunities of AI technologies for qualifications in the future. Our discussions with learning providers have shown that some learners and practitioners are currently using AI tools for a wide variety of purposes, such as for accessibility purposes, translation and generating new ideas for tasks in lessons. Centres have told us that promoting a broad understanding of how AI tools can be used appropriately and effectively amongst their staff and learners could become increasingly important, because of their potential for enhancing learning and professional practices.

Our discussions with learning providers have also included ways in which the content of qualifications may include relevant AI technologies in the future:

New dimensions to existing areas of study: Qualifications in computing, digital, engineering and other areas which are highly technologised may see additional areas of study, relating to AI, form part of their corpus in the future, providing further breadth or opportunities for specialisation. 

New disciplines which relate to new or emerging occupational roles: Many qualifications, often taken by work-based learners, relate to defined occupational roles, which are sometimes based on recognised national standards. These qualifications often allow learners to demonstrate their competence in a specific occupational role. As AI practices develop and become embedded, qualifications may emerge, perhaps in roles relating to the design, implementation and maintenance of AI-enabled systems. 

Reflecting updated practices and ways of working: The use of AI technologies may influence and shape professional and disciplinary practices in a wide range of subjects and domains. Over time, these may be reflected in the content of qualifications as awarding bodies ensure they remain valid for learners taking them. 

We are also engaging with software developers and digital solutions organisations to better understand the ways in which AI tools are being developed and trialled in the context of educational assessment. In doing so, we are considering whether such applications could benefit learners, and whether they could be applied in ways which would be predictable, dependable and explainable.

4. Our next steps

1. We will engage with awarding bodies about their uses of, and ambitions for, AI technologies within their qualifications offers, encouraging them to contact us with any arising questions about compliance with our Standard Conditions of Recognition.
2. We will continue our work on remote invigilation so that we remain up to date with awarding bodies’ use of AI when delivering assessments in this manner.
3. We will engage further with learning providers to remain up to date with their experiences of using AI in teaching, learning and development, relevant to the content of qualifications.
4. We will explore emerging applications of AI for different purposes in educational assessment, evaluating the extent to which they could offer benefits to qualifications, and the qualification system, in Wales.
5. We will continue to engage closely with Welsh Government and other qualifications regulators on the use of AI in qualifications and assessment.