Staff FAQs on the use of GenAI in teaching, learning and assessment

The School’s three-position framework remains the same (no authorised use*, limited authorised use, or full authorised use). Each department is expected to agree its own position at the department or individual course level in advance of the start of each academic year.

Additionally, ‘observed assessment’ is required across all programmes. This means including in-person exams, oral assessments, in-class components or appropriate formative assessments to verify that students are achieving learning outcomes at programme-level. For additional help, see FAQ below on observed assessment options.

In general, you are encouraged to:

• Teach students about effective and ethical GenAI use
• Use LSE-approved GenAI tools (Claude, Copilot) to assist your teaching and marking as appropriate
• Design assessments that integrate GenAI appropriately

All departments have adopted an overarching position on the authorised use of AI in assessment. However this decision might be to delegate to programme directors or course convenors. Check what this is with your , Deputy Head of Department for Teaching/Education, Department AI Lead or Department Manager.

If the decision is yours, consider Table 1 below which offers a brief comparison of the three positions and consider the questions that follow.

Table 1 - understanding the three positions

If the position has been delegated to individual programmes or courses, the following questions may help you to decide on what position to adopt:

1. What skills and knowledge are you assessing?

2. Can you tell if students achieved the learning outcomes with your current assessment methods?

3. How might your students need to work with GenAI professionally, and what foundational skills might you help them develop?

4. Can you clearly explain boundaries to students? (see FAQ below on communicating your approach to students).

If you are adopting a programme-level approach, programme directors will need to engage in discussion with course convenors to agree a whole programme approach which addresses the requirement for observed assessment (See FAQ below on observed assessment).

Need further guidance? Contact your department AI lead or Eden Centre departmental adviser. 

Students say that they are confused as they're taking courses that have a range positions on the use of GenAI within the same programme, and they struggle to keep track of what is permitted for particular courses.

Without clear, consistent communication, students may assume all courses have the same rules, they may struggle to remember where the information about your position is when they need it or there is a risk of accidental misconduct through confusion, rather than intent.

The most effective policy is to make your position impossible to miss, so here are some places you can share it:

• Course Moodle page
• Course outline/syllabus
• In class - provide an explanation with examples and Q&A
• On every formative and summative assessment brief that provides specific guidance when they're actually doing the work.

To help staff, the Eden Centre has published effective communication guidance with templates that are based on and draw upon departmental good practice. Using these will save you time and provide you with ideas.

Resource: Good practice guide: communicating with students about Generative AI in formative and summative assessments 

Observed assessment is an assessment method which reliably confirms that individual students have achieved the work themselves and demonstrated the intended learning outcomes. This may apply to both formative and summative assessment.

The School approved this policy in June 2025 because GenAI use cannot be reliably detected in some assessment methods, such as unsupervised take-home coursework, meaning learning cannot be reliably verified. Observed assessment provides that verification.

LSE requires some elements of observed assessment across a programme. Current sector practice varies from 25% to 60%. Check with your programme director – you may already meet the requirement.

Methods that count as ‘observed’ include:

Linked formative and summative work

• Students submit preparatory stages (e.g. annotated outlines, research proposals, draft sections) allowing you to track their development and verify that summative submissions authentically represent their learning journey.

In-person

• Closed-book invigilated exams

• Oral assessment (presentations, vivas, interactive Q&A)

• In-class work (quizzes, problem-sets, simulations and short-answer test).

Technology-enhanced:

• Edit-tracking platforms: Having engaged in proof-of-concept testing in 2025/26 with the Cadmus assessment platform, LSE remains interested in edit-tracking and authorship tools as potentially useful additions to GenAI and assessment. This year, the Eden Centre will be undertaking a small-scale pilot of the Cursive authorship tool to further build knowledge and understanding of the utility of such tools. 
• Moodle’s quiz settings: these can be optimised to reduce the opportunities for GenAI misuse. For example, set time limits to reduce the likelihood of students looking up answers using GenAI; release correct answers only once the quiz has closed which prevents sharing GenAI generated answers with classmates who haven’t yet taken it; randomise questions from a question bank so that each student gets a unique version of the quiz; show one question at a time to make it harder to collate all the answers at one time; restricting the number of attempts to do the quiz makes it harder to test answers using GenAI and re-enter answers

• Video presentations: using Zoom or Echo360, students record themselves explaining their work, analysing a case, or solving a problem on camera. Examples may include explaining dissertation findings, talking through a coding solution, or applying theory to a new scenario. You may also pose a random question as part of the assessment that require students to respond or ask them to produce a short 1-2 min video explaining what students have written. These aspects can support verification of authenticity. 

Hybrid:

• Written work followed by oral defence

• Staged assessments that combine written and oral components

• Formative activities directly linked to summative grades.

To help you choose what might work best for you and your students, consider what you are really assessing and how this relates to the learning outcomes. Your choice may depend on practical constraints such as cohort size or teaching resources.

If you require further guidance, contact your departmental AI lead or Eden Centre departmental adviser.

Resource: Observed Assessment Policy and Implementation

LSE’s three position policy allows for formative and summative assessments that integrate AI use. Carefully designed assessments allow students to develop their AI fluency and demonstrate critical, creative and thoughtful AI use. Some examples include:

• AI artifacts: students use AI to create multimedia and digital outputs such as podcasts or videos, visualisations and explainers, interactive HTML and websites, chatbots or agents, comic strips, stories or poems.
• An assessed audit trail or process documentation accompanying an assignment: students document their use of AI, alongside other sources, explain their reasoning for using each and triangulate the outcomes. This makes their learning process transparent to both students and markers.
• A critique or comparison of AI outputs: students evaluate AI-generated responses, identifying flaws, biases, or omissions to assess their critical judgement rather than production.
• AI as a starting point: students begin with an AI-generated response and are assessed on the revisions they make to it.
• A dialogue with an AI chatbot: students complete a conversation with a chatbot, for example a role-play in which they must defend and discuss their ideas; they submit their chat log along with a written reflection critically analysing their own choices and the chatbot's performance of its role.

With GenAI widely available, many traditional assessment methods can no longer reliably verify student learning. As detecting the use of GenAI is difficult, better assessment design is one way forward.

How to Build AI Resilience in Assessment is a resource produced in March 2026 by the School’s working group on AI and Academic Integrity (a sub-group of the LSE Department AI leads group, convened March–November 2025). It sets out what AI resilience is, why it is often valuable even on courses where some AI use is authorised, methods of AI resilience and considerations for choosing the approach that is right for individual assessments, courses and programmes.

Assessment designs that help make genuine learning visible and verifiable include:

1. In-person assessment. Written exams, oral exams, presentations and in-class quizzes.

2. Edit tracking. Students write within a specialised platform, such as Cadmus, that records edits as they work. Version history in Google Docs or OneDrive offers a partial alternative.

3. Vivas. After submitting their work, students are invited to discuss it. Interviews may be graded or ungraded.

4. Supervised assessment. Regular meetings to discuss work in progress.

5. Tasks AI Cannot Complete (Yet): Work drawing on knowledge an LLM cannot access (an in-class discussion, for instance), media that AI cannot currently produce, or projects involving face to face interaction such as placements and client-facing work.

For further information visit the LSE Assessment and Feedback Toolkit.

The Eden Centre has also designed ADA – a GenAI tool written in Claude to help staff through the process of designing assessments. Contact Dr Yang Yang via y.yang170@lse.ac.uk if you wish you use this tool. 

Oral assessment is used in undergraduate and postgraduate programmes in many countries. The UK HE sector has tended to rely heavily on written work to demonstrate learning outcomes. Whilst there are clear benefits of demonstrating evidence of learning through writing, this form can be vulnerable to GenAI.

LSE therefore is encouraging the careful introduction of oral assessment as part of its observed assessment policy.

Guidance on designing and delivering oral assessments has been developed by the School’s working group on Oral Assessment (a sub-group of the LSE Department AI leads group, convened March 2025–June 2026).

The following points offer some initial guidance around the process of designing an oral assessment:

• Start by aligning your approach with course learning outcomes and being clear about what you're assessing through the use of making criteria (which should be published in advance so students know what to expect).
• Typical oral assessments last 10-20 minutes per student and can be conducted with individuals or groups. Student preparation is crucial to reduce anxiety and ensure fair assessment.
• Address inclusivity and accessibility early in your assessment planning. Focus your assessment on students’ knowledge and understanding rather than presentation skills or language fluency, which is especially important for students with English as an additional language.
• Record the assessment (audio or video) for quality assurance purposes.
• For grading purposes, use detailed marking criteria and complete your marking during or immediately after the assessment.

Resource: Assessment and Feedback Toolkit – Oral assessments.  

Dissertations represent students' deepest engagement with their subject, typically counting for a quarter of the students’ overall programme and representing months of research and writing. This makes them high-stakes assessments that are particularly vulnerable to cognitive offloading to GenAI. However, there are safeguards you can introduce to verify the authenticity of the individual’s work without altering the grading structure.

• Introducing oral components provides direct verification of student understanding. You could consider mini-viva examinations where students defend their work, or require research proposal presentations and progress presentations at key stages. The resulting conversations reveal students’ understanding of their research. Progressive milestones make the research journey visible and motivating and encourage students to see their assessment as a process rather than just an end product. Requiring formative submissions at key stages such as annotated outlines, comprehensive literature reviews, or chapter drafts that receive supervisor feedback creates a documented trail of development that emphasises the evolution of ideas.
• A structured supervision process would use timetabled supervisory meetings as more formalised opportunities for supervisors to use submitted work as prompts for discussing with the student the development of their thinking. These meetings can give supervisors insight into students’ understanding and help to authenticate their ownership of the work.
• You may also ask students to keep research journals or reflective logs documenting their thinking process, or require primary source evidence such as interview transcripts or survey data. Bibliographies that include personal reflections on source relevance also help to verify genuine engagement with the literature.

For each of these options you can consider allocating a portion of the overall marks to the different components or a single mark made up of the written dissertation and the individual components. 

This kind of enhanced supervision also strengthens your relationship with your students and provides them with valuable learning and feedback opportunities.

Resource: Assessment and Feedback Toolkit - Dissertations.

Yes, you can use interviews to verify student understanding and maintain academic standards. LSE regulations permit two distinct types of interview, randomised and selective, each serving a different purpose.

Both approaches are permitted under LSE regulations and help maintain academic standards while deterring misconduct: 

• for randomised interviews, you must communicate your intention to use interviews in your course outline from the very start of teaching, so students know from day one that they may be asked to discuss their work. 
• for selective interviews arising from concerns about a particular submission, these form part of the evidence-gathering process and do not require prior notification in the course outline. 

More information: Plagiarism guidance for staff.
Guidance Notes for investigating suspected use of Artificial Intelligence (AI) software in summative assessments.

Yes, you can use GenAI in your teaching, provided you follow appropriate safeguards around data protection and transparency. This academic year, LSE provides two approved tools for staff and students: Claude (Anthropic) and Microsoft Copilot. Both offer Commercial Data Protection, meaning your prompts, responses and any uploaded files are confidential and cannot be used to train the GenAI models. This gives you and your students the highest level of data protection available. The School encourages you to use these tools to enhance your teaching.

Copilot is available through your LSE IT account at copilot.microsoft.com. Claude needs requesting before you can use it: https://link.lse.ac.uk/requestClaude

You can also use these tools to develop course materials, generate examples, explore how GenAI might be used in your discipline and to enhance feedback and the quality of marking (see FAQ below on marking and feedback).

If you choose to use platforms other than Claude or Copilot, please be aware that free versions typically don't offer the same data protection guarantees. You and your students’ prompts and work may be used for training, and privacy protections vary significantly. There are also significant differences in capability between free and premium versions of most LLMs, including access to more advanced models, additional tools and higher usage limits. If you have only used the free offerings, we encourage you to experiment with the premium version of Claude to see how it compares. For example, try using 'Research' mode, which is not currently available on free accounts.

Important note: Using GenAI in your teaching doesn't change your department/programme or course position on student use of GenAI. You still need to communicate clearly to students what they are permitted to do (see FAQ above on communicating your position).

We encourage full transparency with students about your use of AI in the development of teaching and learning materials as a means to building trusting relationships with them and encouraging reciprocal transparency.

For further information, visit Copilot at LSE or Claude for Education at LSE.

Resource: Supporting LSE educators to use GenAI.

Understanding Claude: https://lsecloud.sharepoint.com/sites/DSLLearningHub/SitePages/Academic-Student-Claude-homepage.aspx

Understanding Copilot: https://lsecloud.sharepoint.com/sites/DSLLearningHub/SitePages/Staff-Copilot-homepage.aspx

There are many practical ways to introduce GenAI meaningfully into your classroom.

Start by building shared understanding with your students. Consider co-creating ground rules on using GenAI together, including how to cite GenAI appropriately if its use is permitted in your assessments. This collaborative approach helps students understand the reasoning behind your position and gives them ownership of the guidelines.

Model critical use. Demonstrate both effective and problematic uses of GenAI in your discipline, pointing out where tools produce hallucinations, errors, or misleading outputs. Encourage students to ask questions about methodology, attribution, citations, authorship, and ethics as you work through examples together. You might also use GenAI collectively to build understanding of a topic, explicitly noting the strengths and weaknesses of GenAI-generated content compared to traditional scholarly sources. Promoting discussion, modelling tool use, and creating shared understandings all help reduce student anxiety while enhancing their GenAI fluency. Students learn to use these tools critically rather than accepting their outputs without evaluation.

Create custom learning assistants for your course. Using Claude Projects or Copilot, you can develop teaching assistants that have access to the sources and literature used in your course. These custom tools can support student learning by following your instructions about how to behave while remaining grounded in your course materials.

Design simulations for experiential learning. You could use GenAI to role play scenarios such as a client negotiation or dialogue with a historical or political figure. Ensure they debrief with you and/or each other after the interaction to explore their learning and reflect on their reasoning processes.

Scaffold learning a complex process or concept. There are various creative ways for your students to approach this:

• Working with the AI to develop concept maps;
• Asking the AI to explain a concept to a variety of different audiences or through different analogies;
• Asking the AI to make a deliberate error in its answer and then finding and correcting the mistake;
• Asking the AI not to give the answer but to ask them questions of increasing complexity.

If you are interested in developing learning activities that integrate AI, please contact Steven Williams or Jamie Hayward,Eden’s AI and Education Specialists via s.williams10@lse.ac.uk and j.hayward1@lse.ac.uk.

Resource: Practice and perspectives from LSE

LSE is committed to using AI to enhance rather than replacing student reasoning, to support critical thinking, structured enquiry, and the development of AI literacy. Surfacing and sharing examples of colleagues’ effective practice in this area provides a network to celebrate, disseminate and evaluate best practice and foster an engaged community. The Eden Centre does this through the ongoing development of events and resources, including:

• The AI and Education Showcase
• The annual Education Symposium and Forum
• CPD workshops
• Case studies and videos of practice
• The AI and Education Toolkit

Students need to understand both the practical limitations and broader ethical implications of GenAI tools to use them responsibly in their academic and professional lives.

GenAI-generated content may contain errors, biases, or completely fabricated citations and references (hallucinations). Students who over-rely on these tools risk undermining their own deeper learning. They need to understand that while GenAI can support their work, they remain fully responsible for accuracy and must put in genuine intellectual effort. The learning and the work still needs to be theirs.

Beyond concerns about accuracy, GenAI also raises significant ethical issues that include substantial environmental impact through increased energy and water consumption, problematic labour practices in training GenAI models, widespread use of copyrighted materials without permission, perpetuation of societal biases and stereotypes, and growing inequalities between those who can afford premium GenAI tools and those who cannot.

In class, you can open up discussions about these issues and model responsible GenAI use yourself. Show students how GenAI can support rather than replace their thinking, demonstrate the importance of verification and fact-checking, and model effective prompting to help students use tools efficiently. Discuss the specific limitations and applications of GenAI in your discipline and explore how GenAI is currently being used in professional contexts in your field, including any relevant professional or ethical standards that apply.

Resource: Supporting your students

There are many ways that GenAI can help you with feedback and enhancing quality assurance processes. For example, converting handwritten notes to typed feedback to improve accessibility and legibility; analysing feedback patterns across a cohort to reveal common student challenges and inform your teaching; and checking feedback clarity to ensure students understand your guidance.

In collaboration with King's College London and the University of Southampton, LSE has developed AI-assisted marking and feedback principles and accompanying guidance and scenarios that illustrate the necessary safeguards to assist you in navigating this emerging practice.

What remains your responsibility:

• You must read and assess all student work yourself. You make all marking decisions independently - GenAI supports your process but never replaces your judgment.
• You maintain oversight at every stage and remain transparent with students about GenAI use.
• You avoid uploading identifiable student work without explicit consent.

GenAI cannot determine marks, replace second markers or moderators, or make final assessment decisions. These remain entirely with academic staff.

The Use of GenAI-assisted marking and feedback at LSE is a set of principles and also includes seven practical scenarios: scaling feedback for large cohorts, supporting staff accessibility needs, digitising handwritten annotations, calibrating feedback tone with grades awarded, identifying teaching improvements from feedback patterns, checking whether feedback makes sense to students, and supporting moderation calibration.

These examples show GenAI enhancing marking practice within appropriate boundaries.

Resource: Use of GenAI-assisted marking and feedback at LSE

LSE has published updated guidance for staff on how to conduct investigations into AI-related academic misconduct. It sets out a six-stage framework covering the appointment and training of investigators, the initial review of flagged work, an expanded review of the student's other submissions, the conduct of student interviews, evidence evaluation, and the formal allegation and penalty process. The document identifies various forms of evidence that may indicate AI use, including confabulated references, stylistic inconsistencies, and hallmarks of AI-generated text. It also cautions against over-reliance on AI detection tools. The guidance emphasises that protecting students from false accusations is paramount, while also clarifying that a student confession is not required where evidence strongly indicates misconduct. 

The preventive approach remains most effective. Rather than trying to detect GenAI use after submission, design assessments that verify learning directly through observed methods, progressive checkpoints, or formats that are more resilient to GenAI misuse or that make unauthorised GenAI use more obvious (see FAQ above on observed assessment and FAQ above on re-designing assessment) and engage in discussion with your students about the effective and appropriate use of GenAI. 

More information: Plagiarism guidance for staff, including guidance notes on AI misconduct.

How you handle student use of GenAI in assessment depends on the position you have adopted and communicated to students.

• If you've adopted Position 1 (No Authorised Use), the boundaries are clear. Undocumented GenAI use constitutes potential academic misconduct and should be handled through proper channels. When you encounter work that raises concerns, mark it on its academic merit first, then flag any integrity concerns separately through the standard misconduct procedures.
• For Position 2 (Limited Authorised Use), you need to check how students have documented their GenAI use against the instructions given. If students have documented their GenAI use properly but appear to have relied on it too heavily, this is an academic quality issue rather than misconduct. Mark the work accordingly, recognising that over-reliance on GenAI typically results in weaker demonstration of original thinking and lower marks.
• With Position 3 (Full Authorised Use), students have significant freedom but must still document their GenAI use. When GenAI is used appropriately with proper documentation, mark the work normally on its academic merits. If the work shows little original thought or critical engagement despite technical proficiency, mark it accordingly – this is an academic quality issue. However, if students fail to document their GenAI use at all, this missing documentation may still constitute misconduct as it violates the transparency requirements you've established.

LSE’s key principles remain constant across all positions: mark the work on its academic merit first, assessing how well it demonstrates the learning outcomes and meets the assessment criteria. Address any integrity concerns separately through proper channels. This separation ensures fair academic assessment while maintaining appropriate processes for potential misconduct.

Where can I get help implementing these policies?

Assessment and pedagogy support:

• Eden Centre departmental advisers;
• Eden Centre resources and workshops; 
• Email: eden@lse.ac.uk.

Technical and administrative:

• TQARO – assessment changes in Akari via tqaro@lse.ac.uk;
• IT Services – Claude and Copilot technical support via tech.support@lse.ac.uk;
• Digital Skills Lab – using AI tools effectively via j.l.brown@lse.ac.uk.

Policy and regulations:

• Contact your departmental AI lead
• Contact the Academic Registrar's Division for advice on misconduct procedures
• Contact the Disability and Mental Health Support team to find out more about reasonable adjustments