Policy for the use of AI/LLM

For students For teachers

General Principle

The use of AI and large language models (LLMs) is permitted in the MSc Geomatics for the Built Environment programme as a learning aid for writing, coding, research, and general exploration. AI tools can help you learn, but they must not replace your learning.

	Task	Details
✅	Brainstorming / inspiration	No disclosure needed unless AI substantially shaped the submitted content
✅	Explanation of concepts	Using LLMs as a “Google on steroids” to help you understand complex concepts
✅	Grammar / style / language checking	Tools like Grammarly are treated as general aids; disclose if significant rewriting occurred. Example: asking an LLM to write a 500-word discussion section for your report is substantial; asking it to rephrase one awkward sentence is not.
✅	Translation of texts	Disclose tool in reference list if used for translating academic work
✅	Summarising existing literature	Always verify AI-generated summaries against original sources; cite the original, not the AI. Making a summary podcast from complex material is also allowed. You are not allowed to submit an AI-summary as your own work.
⚠️	Drafting text submitted as assessed work	Check course-specific guidelines on LLM/AI use for additional requirements. Complete disclosure required — see rules below
⚠️	Code generation (submitted assignments)	Check course-specific guidelines. If allowed, disclose and be prepared to explain every line in an oral authenticity check
❌	Using AI during a closed written exam	Prohibited unless the examiner explicitly permits it and announces this before the exam starts
❌	Submitting AI-generated code, text, and images without disclosure	Counts as fraud — see below

Your responsibility

As a student, you are personally and fully responsible for all work you submit, including any content produced with (the help of) AI/LLM. Concretely, this means:

• You must be able to explain, defend, and reproduce the substance of any submitted work.
• You must critically evaluate all AI/LLM-generated output. AI tools can produce plausible-sounding but incorrect, biased, or fabricated information.
• You are expected to adhere to the principles of academic integrity of TU Delft:
  • TU Delft Integrity Portal
  • TU Delft Code of Conduct
  • Netherlands Code of Conduct for Research Integrity

Per-course rules

Each course will clearly state in its course guide what extent of AI/LLM use is permitted for assignments, coding, writing, and research activities. Some courses will allow more than others:

• Q1–Q3 courses tend to be more restrictive, as these cover foundational concepts that you must master yourself.
• Later courses may allow broader use of AI tools.

When in doubt, ask the instructor before using AI.

If a course does not specify its own rules, the default rules of this document apply.

When AI use becomes fraud

The core test is: does the AI output replace the knowledge or skills you are being assessed on? If yes, it is fraud.

The following are considered fraud:

• Submitting AI/LLM-generated text, code, or analysis as your own work without disclosure
• Letting AI produce the core deliverable of an assignment (e.g., AI writes your report, solves your problem set, or generates your code solution)
• Using AI during exams or assessments where it is not explicitly permitted
• Using AI to paraphrase or rewrite sources to avoid proper citation

Disclosure Requirements

Whenever you use AI tools in a submitted assignment (and such use is permitted), you must include a clear disclosure statement containing:

1. What: Name and version of the AI/LLM tool used (e.g. DeepSeek V4 Flash, ChatGPT-5.2, Claude Mythos, etc.)
2. How: Purpose and method of use—what you asked it to do and how you used its output
3. Extent: Describe the scope/extent of your AI/LLM use (e.g. which parts of the work involved AI, at what stage, and how many iterations). Give concrete examples.
4. Reflection: Was your use of the AI/LLM tool helpful? Write a short reflection based on your experience.

This section does not count towards a maximum number of pages/words. It can be short for small assignments (max 1/2 page), and perhaps 1 page for longer reports; see course instructions for more detail.

Failure to disclose use of AI when it has taken place is itself a violation and may be treated as fraud.

Thesis and Graduation Project

The thesis is the culmination of the MSc Geomatics for the Built Environment programme, where you must demonstrate that you master the programme’s learning outcomes. Therefore:

• A detailed disclosure statement is mandatory. The MSc Geomatics for the Built Environment thesis template has an Appendix to help you start.
• You must be able to independently defend all scientific content and contributions in your thesis.

Consequences

Suspicion of fraud is sufficient for the teacher to conduct an oral authenticity check to verify the student’s knowledge and authorship of the submitted work.

If the oral check raises further concerns, the teacher will report the case as suspected fraud to the Board of Examiners (BoE) of the Faculty of Architecture and the Built Environment.

The BoE determines whether fraud has occurred and decides on sanctions in accordance with the TU Delft Examination Regulations.

Do ✅

1. Publish a clear AI/LLM policy on your course website (on Brightspace or own website) and highlight where/if it differs from the Geomatics one.
2. Design at least one AI-resilient assessment per course. An AI-resilient assessment requires students to demonstrate process, not just product—for example, defending their choices orally, showing intermediate drafts, or completing a timed component under supervision. Examples of AI-resilient assessments: in-person exam on paper, oral authenticity check, presentation of results, and in-class task.
3. If you suspect fraud, report it to the BoE. You are not the judge; you report facts about your suspicions and the BoE will investigate.
4. For thesis / final project supervision: Plan regular (at least every 2 weeks) interim checks to verify the progress and communicate this to students at the start.

Don’t ❌

1. Use AI-detection software as only evidence of fraud.
2. Automate grading with AI/LLM without review.
3. Use AI/LLM tools to process personal student data. Even for TU Delft’s Copilot it is not certain it is safe (!?). If you need to process student work with AI, ensure all personal identifiers are removed first and use only TU Delft-approved tools (Copilot at the moment).

What to do if you suspect fraud

If you suspect that fraud might have taken place, you must request an oral authenticity check to verify the student’s knowledge and authorship of the submitted work.

If the oral check raises further concerns, you have to report the case as suspected fraud to the Board of Examiners (BoE) of the Faculty of Architecture and the Built Environment. The BoE determines whether fraud has occurred and decides on sanctions, in accordance with the TU Delft Examination Regulations.

If the oral authenticity check raises concerns but does not clearly confirm or refute fraud, you may ask them to do a retake of the submitted assignment.