UTLearning & Teaching PortalL&T NewsLatest update on AI in education

Latest update on AI in education

Developments around ChatGPT and similar AI tools are moving fast. Since the first version of ChatGPT was launched in November last year, the development of this tool has been exponential.  We cannot predict how further developments will evolve, but it is highly likely that the next version would be a similar leap forward for this technology. Solutions like ChatGPT will not go away, but will continue to grow. Therefore, it is important to be prepared for its implications for education and research at UT. In this news item, we have gathered the most important updates.

Recognising AI: not a rat race

From education perspective, it has always been important to be able and adapt to new technological developments. In the long term, there might be a ‘rat race’ from measures and countermeasures in education. For instance, the development of AI which can recognize AI-generated content. We do not believe in such a rat race, and it seems more sustainable to adapt our education (and assessment) in a way that we can deal with the technology. This means that we will have to embrace the technology instead of resisting it.

Important to note here, is that we have to embrace the technology carefully. There are many opportunities but also threats if we are not using AI responsibly, ethically and transparently. Strengthening the human factor in our education seems like the way forward. We already have a culture of academic and scientific integrity combined with curiosity that is cherished by our staff and students alike. We need to support staff and students to apply this mentality to AI and ChatGPT specifically.

Working group on AI in education

For this reason, a dedicated working group for AI in education has been established to adapt our education in the long run. The goals of the working group are to shape the future education forms with the incorporation of AI and to inform staff and students that improper use of AI/ChatGPT is considered academic misconduct.

The working group consists of working group members from various departments and faculties, with experts in a number of relevant disciplines.

To provide a head start, we now have a webpage on AI in education at the Learning & Teaching Portal. Here you can find current information and advice regarding the use of AI in education, with themes such as assessment and what to do with it. AI In Education | Learning & Teaching Portal (utwente.nl)

We would like to stress that it is important to be well aware of privacy and data security when using  this technology. Most of the current platforms process and sometimes even claim the data you use as input and output. Be aware to avoid using any personal, private and/or classified information with platforms such as ChatGPT!

ChatGPT and UT regulations

Earlier, we indicated that we were going to take a critical look at whether our Student Charter is prepared for possible misuse of AI. This has resulted in a further clarification that provides interpretation on how the charter should be read in the light of AI developments and helps the staff of the study programmes deal properly with student use of AI.

This explanatory note reads:

“While embracing the technology, misuse of (generative) Artificial Intelligence applications could be considered fraud. The Student Charter describes what the university considers cheating or fraud. Cheating/fraud refers to any action or negligence on the part of a student that precludes an accurate assessment of the student’s knowledge, understanding and skills (Article 6.6 Student Charter).

Article 6.6 paragraph 1 of the Student Charter states that (any form of) assistance, resources or devices (electronic or technological) other than the ones whose use the examiner or supervisor has permitted prior to the start of the study unit and/or exam or test, or whose use the student knew or ought to have known was not permitted during a test or exam is considered cheating/fraud. Generated Artificial Intelligence programs or applications are considered “assistance, resources or devices” as mentioned in the article referred to above. Consequently, under the current Student Charter, prior permission by examiner or supervisor is needed for the use of generated Artificial Intelligence. A written assignment, project, essay or thesis falls under the umbrella of test or exam.

Article 6.6 paragraph 4 of the Student Charter states that plagiarism is a particular kind of cheating/fraud, which occurs when the student uses someone else’s work or previous work of their own, without correct referencing. This includes, but is not limited to using parts of another text (printed or digital) without referencing (also if minor changes have been made) or using software without referencing. This means that under the current Student Charter, the use of Artificial Intelligence needs correct referencing.

The Examination Board decides whether cheating/fraud has occurred. The Examination Board of the educational programme drafts Rules & Regulations on cheating/fraud. These may include additional provisions to the Student Charter and specify what measures will be taken in cases of (suspected) cheating/fraud.”

Tools for detecting use of AI

In the long run, AI writing detection systems are probably not a sustainable way forward since we will have to adapt to the technology. However, we do like to inform you that the UT has a temporary license for a product. On April 4 Turnitin has released the first version of AI writing detection in their products. This tool is available in SimCheck (in Canvas or the stand-alone web application). In addition to the existing similarity percentage, the similarity report will show an overall percentage of the text that may have been AI-written and highlights the corresponding segments in the text. At the moment this only works for texts written in English. As the AI detection of Turnitin is still in development, we advise you not to use the results as proof against students without further investigation. Just as regular Turnitin scores, this can only be an indication, especially with the number of false negatives and false positives these types of detection software have.

Most likely, this AI detection preview will continue to be available without extra costs until 1 January 2024. If the UT decides that AI detection is a useful addition to the existing plagiarism detection functionality, we will need to upgrade to another license. This will be explored in the upcoming period.

For more information, see this FAQ about the Turnitin AI writing detection.

Latest developments worldwide

As mentioned, there is a lot happening in terms of technological developments. While ChatGPT made quite some impact already, we have seen a lot more happening. The best way to distinguish these technical developments is to split them up into two different movements. The first one is the further innovation and development of the AI models and systems on their own, while the second trend is the integration of these AI models and systems in contemporary ecosystems.

Again, the big names now are ChatGPT and OpenAI: this company released GPT-4 on the 14th of March 2023. While not a huge game changer on its own compared to the introduction of previous versions, the overall quality was vastly improved and allowed for images to be used as input as well. Then there are other AI chatbots rapidly being developed, like Google’s Bard as a competitor. Besides AI models that generate text as an output, there are also the ones that generate images like MidJourney and Dall-E, or sound like OpenAI Jukebox.

While these technologies on their own are being improved upon rapidly, another significant movement is that of incorporating AI models and technology in more contemporary systems. There are big initiatives like Microsoft’s Copilot integrating with the whole Microsoft 365 ecosystem and Bing chatbot adding GPT-4 technology upon a regular search engine. Then there are more confined examples of educational tooling including it. To name a few, Khan Academy introduced its AI assistant to help students and Wooclap’s Quiz Wizard using ChatGPT technology to help generate quiz questions for Wooclap.

More information

For further information/questions regarding the different technologies, you can contact the TELT team. They recently released a podcast about AI in education. EduTalks # AI in Education:   https://share.transistor.fm/s/6d3797d1

Also, a publication of SURF dives a little deeper into the ‘promises of AI in education’, Promises of AI in Education: Discussing the impact of AI systems in educational practices. This report from 2022 delves into the use of AI in educational practices on a micro, meso, and macro scale.  SURF also has a collection of other sources and information on this topic: ChatGPT - verzameling bronnen | SURF Communities

Educational consultant Robin van Emmerloot spoke to regional broadcaster RTV Oost to speak about AI in education (in Dutch).