What exactly are the OfS asking UK universities to do?

The Office for Students (OfS) has published a briefing note setting out institutional practices it considers likely to comply with regulatory conditions B1, B2 and B4, now framed as requiring UK universities to act on AI. The guidance covers embedding AI-related subject developments and transferable AI skills into courses under B1, ensuring quality of AI-generated teaching resources and feedback plus clear assessment-use policies and mitigation of disparities from differential access to paid AI tools under B2, and using secure assessment methods and preserving learning-outcome validity under B4. The note follows a report published the previous day and signals the regulator's expectations for how universities must adapt curricula, assessment and student support to AI.

I’m writing a longer post on yesterday’s report but it’s worth highlighting separately how clear they are being about their expectations. In this briefing note https://www.officeforstudents.org.uk/for-providers/quality-and-standards/artificial-intelligence-in-higher-education/effective-institutional-practice/ they highlight approaches which are likely to be compliant with each of conditions B1, B2, B4 which are now being framed as necessitating action on AI: Approaches likely to be compliant with condition B1 https://www.officeforstudents.org.uk/for-providers/quality-and-standards/artificial-intelligence-in-higher-education/effective-institutional-practice/ collapse-63ca5c6b-118d-41e1-882d-d1d310d598cc - Ensuring appropriate consideration is given to how courses are informed by recent subject matter developments relating to AI. For example: - In computer science, considering the implications of machine learning for approaches to computing. - In law, considering the implications of AI for issues such as personal liability, copyright and data protection. - In education, considering the implications of AI for learning and assessment. - Ensuring appropriate consideration is given to developing AI skills required for success in a relevant industry, such as the ability to understand or effectively use a widely adopted tool in that industry. - Ensuring appropriate consideration is given to developing transferable AI skills, such as the ability to decide when and how an AI tool can support the effective completion of a task or the ability to critically evaluate AI outputs. Approaches likely to be compliant with condition B2 https://www.officeforstudents.org.uk/for-providers/quality-and-standards/artificial-intelligence-in-higher-education/effective-institutional-practice/ collapse-d530a492-227a-4e07-8010-7e81abccea33 - Ensuring that, if AI tools are used to support the development of teaching resources, those resources are high quality. This includes being factually accurate, providing sufficient depth and detail for the subject and level of study, referring students to relevant sources, and being aligned with other learning materials to support a coherent learning experience across the course. - Ensuring that, if AI tools are used to support the preparation of feedback, there is sufficient oversight to ensure this is accurate and aligned to assessment criteria, tailored to students’ work and specific about how they can improve. - Setting out clearly for students whether and how they are permitted to use AI in the process of completing assessments, for example through realistic, meaningful and transparent academic regulations and accessible policy and guidance, and providing clear and consistent support to enable students to understand and comply with any policies on permitted AI use in assessment. - Taking steps to ensure that any disparities between different student groups created by differential access to AI tools, including paid versions, and training are effectively mitigated by the steps the institution takes to ensure its students receive sufficient support for their studies. Approaches likely to be compliant with condition B4 https://www.officeforstudents.org.uk/for-providers/quality-and-standards/artificial-intelligence-in-higher-education/effective-institutional-practice/ collapse-121c3a27-9329-4cfb-a0b9-afcf9079a094 - Using secure assessment methods defined below where these are necessary to ensure assessments appropriately measure whether students have achieved the intended learning outcomes. - Ensuring that, where the use of AI is permitted, assessments continue to require students to demonstrate the intended learning outcomes, which may include the effective use of AI. - Ensuring that, where some AI or other digital tools are provided to students selectively for use in assessments, for example as reasonable adjustments, the assessment remains valid because the tools do not provide an unfair advantage. - Drawing on a range of evidence as part of academic misconduct processes in a fair and balanced way that will facilitate the detection of actual misconduct, rather than placing disproportionate weight on the outputs of AI detection tools. We use the term secure to mean assessments which are designed to prevent the use of AI in the assessment process. They include in-person examinations, presentations, vivas oral examinations , performance-based assessments, and assessments that require students to demonstrate their knowledge and skills through practical or applied tasks. In addition to these considerations relating to assessment and AI, universities and colleges also have a legal duty not to discriminate against students. This may require them to make reasonable adjustments for disabled students to remove a substantial disadvantage that results from their assessment processes. This may include adjusting assessment methods where necessary to ensure the assessment appropriately measures whether students have achieved the intended learning outcomes. In the broadest sense I think this is asking universities: - Secure assessment in robust but equitable ways - Communicate expectations about AI use clearly to students - Incorporate AI-related skills and content into curricula - Handle suspected misconduct fairly, rather that through detection - Action to reduce disparities of access to AI tools - Oversight of the use of AI tools by staff and in processing assessment That is… not unreasonable, to put it mildly. Indeed one might ask why the sector is so far from this being in place nearly four years after ChatGPT was launched.