AI Guidance at the University of Glasgow

The University’s position with respect to students’ use of generative AI tools continues to be that we recognise that these technologies are changing the world of work and that, consequently, we consider it essential that students develop the knowledge and skills to make appropriate use of these resources and technologies during their studies.

Staff Support and Resources

Academic colleagues need to determine how best to advise students on the use of generative AI tools in the context of the courses that they teach and the assessments that they design. As previously advised, the best course of action is to re-design assessment in a way that allows for appropriate use of all relevant resources and tools as proposed in the Learning Through Assessment Framework (LTA). Development of staff resources to support colleagues in making such changes is continuing and new resources will be uploaded the Assessment & Feedback Resources Hub on an ongoing basis.

It is recognised that redesign can take time for colleagues to undertake and for LTA to become embedded so resources and support will continue to be provided over the longer term. Policy and practice guidance will also be continually reviewed and updated given the fast-paced nature of technology development and the associated changing environment.

School Communications with Students

For each component of assessment, there is a need for students to understand clearly how they can and cannot use generative AI. In determining the approach, the scenarios below should be considered, and actions taken accordingly, well in advance of students beginning to engage with assessments (both formative and summative). It is expected that in most cases, a School statement, embedded in relevant course documentation, will be feasible and desirable.  It is also expected that the majority of cases will align with Scenario 1. School L&T Convenors are asked to coordinate communications across the School to ensure that such information is available to students ahead of them beginning their studies. Deans of Learning and Teaching will collate those updates through College L&T Committee structures.

Scenario 1: No assessment-specific requirements in terms of generative AI usage

Where colleagues have no specific requirements in terms of generative AI usage, the default assumption will be that students are likely to make some use of generative AI when engaging with their assessments. Schools should signpost the guidance being prepared by Student Learning Development which supports students to make appropriate use of such technologies in a manner that is consistent with our policies on academic practice.   Colleagues are strongly encouraged to talk to students about how to effectively use resources for their course assessments, and where possible provide demonstrations.  Example approaches could range from providing advice on how to reference or evidence use of AI generated content, to building in AI use as part of the assessment preparation and to using AI generated output as part of the assessment itself. Some specific examples illustrating the kinds of student guidance that would be appropriate to these approaches are given below but there will be a range of possibilities across the University.  The key message is to make sure that students are in no doubt of what is expected and acceptable.

  • Example 1: On this course, students are permitted to use any AI generated content that they have used, though must fully attribute it in the main body of their work. When indicating this to students, colleagues should note that currently, many referencing styles do not provide specific instructions on how to cite AI tools. In order to ensure avoiding academic misconduct, AI should be referenced – using the conventions within each subject – to provide specific details on: originator of the communication (e.g., ChatGPT/OpenAI, Google Bard, etc.); the exact title/details of the prompt(s) used; and the date of use. Current best guidance is to treat referencing any AI tools as a form of personal communication, and to follow the referencing style’s conventions for that form of reference.
  • Example 2: On this course, and to develop your critical thinking skills, you are asked to generate a draft essay using Chat GPT or similar, and to provide a critique of that output reflecting on the extent to which a range of appropriate resources have been cited, evaluating whether a suitably broad range of perspectives have been considered etc.
  • Example 3: During this course you will be given the opportunity to discuss as a group a Chat GPT response to a question about Topic X, and we will explore the features of that response and thus the academic limitations of such technologies etc.

Some short videos are provided illustrating how colleagues could consider adapting assessments in ways that recognise use of Chat GPT and other similar technologies. 

 Please ensure that, in assessment guidance, there is direct reference to student-facing guidance on appropriate use of academic and other resources. 

Scenario 2: Generative AI can be used only in a particular way

If it is determined that generative AI use is permitted in particular ways for a piece of assessment, then colleagues should ensure that they provide clear instruction to students in advance of the course beginning, so that students understand what they can and cannot do. Instructions need to be highlighted to students, and provided appropriately within course Moodles and/or handbooks along with other learning, teaching and assessment guidance and with specific reference to relevant assessments. In such cases, it is important to bear in mind that AI detection tools remain unreliable and imperfect, with a significant likelihood of yielding false positive or false negative results, and are therefore not being adopted by the University. Hence, careful consideration must be given to any limitations placed on students’ use and how any deviation from that could, realistically, be identified and addressed.

Scenario 3: Non-use of generative AI

As we move more and more to meaningful, iterative, programmatic and inclusive assessment, the reliance on invigilated, time-limited examinations is expected to decrease and not increase. This direction of travel is in keeping with guidance from quality bodies (such as QAA). However, we recognise that it can take time to re-design assessment. Thus, where it is considered that there is no alternative assessment design in the short-medium term, then we will apply the criteria that we have applied in previous exam diets to make and prioritise decisions around hosting exams on campus. Arrangements for examinations will be made in the usual way through discussion with School L&T Leads and Deans of Learning and Teaching.