Fraud / Generative AI (Chatgpt)

How do you deal with the latest developments around generative AI like ChatGPT in (higher) education? Can you still conduct reliable assessments if students can get the information and whole texts ready-made? Can we prevent students from using these kinds of tools? Should we? Is it to be seen as fraud? What policy is needed? Should we want to prevent it or can we actually make good use of it? At UT (and elsewhere), the risks but also the added value of this development are being explored. 
Below are a number of sources, both from the UT itself and from outside the UT related to this topic. 

  • Check out the latest news on the AI in Education website (UT)

    On this site regular posts and the latest news regarding AI and Education (at the UT) can be found. 

  • Newsletter from the working group AI in Education at the UT - 7 July 2023
  • What can be used to detect the use of AI?

    The developments regarding AI are going fast. Detection has become kind of a cat-and-mouse game. No AI detector can provide complete accuracy, especially as language models continue to evolve, with every chance for false positives and negatives. 
    Important points to consider are also privacy and security issues. If you use free tools, chances are that they might compromise user privacy or security. When paid services are used, data protection and compliance with privacy regulations might be better, but this has to be well established.  
    For the UT emplyees we now have Microsoft Copilot available. On the Copilot website you can work with this chatbot by using your UT-account. But although it is a closed-off environment, it does not (yet) adhere to the privacy agreements in data processing for the UT and any confidential data like personal, research or student information, should not be used [read more in Available for all employees: Microsoft Copilot (utwente.nl)]. Nevertheless Co pilot can be useful and assist you in your education efforts in other ways, see Meet your AI assistant for education: Microsoft Copilot | Microsoft Education Blog

    So for detection it might work better to trust your own eyes and feelings, although to prove something remains difficult. As teacher you can be alert on 'signals' suchs as:

    • (In)Consistency in Style. Long, complex sentences. Lengthy elaborations. 
    • Use of specialized terminology that differs from the terminology used in the course.
    • Mastery of the content exceeds what was offered in class. 
    • (In)Consistency in the quality of work across assignments.
    • Incorrect references. Unusual references. 
    • The work of the student doesn’t follow the required document formatting.
    • Similarity of work if a lot of stduents use AI with the same prompts. For instance for a programming or coding assignment. 
    • Unusual punctuation marks in a text. 

    You can also try out what will happen if you put your assignment text in an AI programme or the prompts you expect your students will use. 
    An interesting article about this topic (commercial site): Universities can detect ChatGPT, at least to some extent, by Funmi Somoye, in PC Guide (April 2024)

    Some detection tools mentioned by AI self (but keep in mind what is said in the introduction): 
    GPTZero // Scribbr’s AI Detector // Turnitin // CopyLeaks // Undetectable.ai. In this article some other tools are named: GPT detector
     

University Twente sources

SURF sources (Netherlands)

SURF is a cooperative association of Dutch educational and research institutions. Their aim is to work together to acquire or develop the best possible digital services, and to encourage knowledge sharing through continuous innovation. SURF published several articles and blogs about ChatGPT and offers webinars about this theme. Check out their website for recent activities, like: a publication of SURF Promises of AI in Education(2022) and a collection of other sources and information: ChatGPT - verzameling bronnen | SURF Communities.  Artificial intelligence en machine learning | SURF.nl and see some examples below: 
How big can the impact of Language Models on Education be? | SURF Communities || From Eliza to ChatGPT: the stormy development of language models  | SURF Communities ||  Video - Let's have a (human) chat on ChatGPT by Thomas van Osch | SURF Communities 

National sources

International sources

Courses

There are courses on this topic available free online. Below some are mentioned (not checked for quality), but there surely will be more.
* If you follow one or know about a useful one for teachers, just let us know!