There are many exciting developments happening in the field of Artificial Intelligence (AI). What implications does this have for education at Erasmus University Rotterdam (EUR)? On this page, you will find an overview of relevant information and developments within our university. This information has been gathered from contributions by experts, users, and task forces. AI has been on everyone's mind for a while now, with developments following each other at lightning speed. For this reason, this page will be updated regularly.
Generative AI
Recently, generative AI has received a lot of attention. This is a branch of AI that focuses on developing algorithms and systems capable of creating new content, such as images, music or text. Unlike traditional AI systems that are programmed to perform specific tasks, generative AI algorithms can learn from user input and independently generate content similar to human creations. This is done, for example, through machine learning, deep learning or neural networks. Generative AI has many applications, including the development of chatbots, video games, virtual assistants and more. ChatGPT is now a well-known example of this, but numerous other examples can be found.
Practical information for lecturers
The information under this heading was collected by the ESE Learning and Innovation Team and the EUR task force AI. Given the rapid pace of developments, we cannot guarantee that information is always fully up-to date. Updates will be made when necessary. If you have any suggestions or updates to this page, please email the Community for Learning & Innovation. For any school-specific questions, please reach out to your Learning & Innovation team.
Generative AI can impact education in many ways. Currently, the technology is still in its infancy, and it is clear that schools cannot yet make a long-term estimate of the potential benefits and downsides of using generative AI in their courses. Therefore, some caution and restraint are appropriate before deciding to adopt AI tools in our education. However, these tools are available, and we must operate under the assumption that students are using them. In the short term, we see mostly a threat for assessment.
⚠️ Be aware that generative AI tools are provided by commercial entities. Any data you enter is being stored and can be used to further train the model. Therefore, you should not upload any private, sensitive, or confidential material. This also means that you cannot force students to use any of these tools. If you decide to adopt generative AI in your course, you will always have to provide an alternative to students.
- Generative AI tools like ChatGPT and Bing Chat will have an impact on our education.
- Currently, the biggest impact of generative AI is on our ability to asses assignments. If your course uses assignments, please use the AI assessment guidelines within your own faculty/school.
- If you want to integrate tools like ChatGPT into your course, please reach out to your Learning & Innovation team.
In the short term, we recommend that most courses do not significantly change their content or set-up in a response to generative AI. In making this recommendation, we also recognize that for some courses generative AI could be an excellent tool to improve education. For these courses, it may be beneficial to actively integrate generative AI in teaching or assessment.
Irrespective of your decision to integrate generative AI into your course, or not, some active involvement on your side is required. For any questions, please reach out to your Learning & Innovation team to discuss the best way forward.
Within each faculty, the examination board has an important role in dealing with and preventing fraud. An examination board is an independent body within the faculty that focuses on monitoring the quality of education. When a student uses AI software such as ChatGPT without permission from the examiner (the person who develops a course and its exam), this is considered plagiarism or ghostwriting by the examination board. Appropriate sanctions may follow. The specific sanctions depend on the circumstances and the policy of the faculty concerned. Each faculty and each case is therefore handled differently.
There are three ways in which students can misuse generative AI:
- Generative AI is not allowed, but the student still makes use of these tools. You can compare this to ‘ghostwriting’.
- Generative AI is allowed, but the student uses it in the wrong way. For instance, the student does not explain and reflect on the use when you made that a requirement. You can compare this to wrong references. Keep in mind that referring to generative AI is new for students. Mistakes may happen.
- Students use generative AI for ‘traditional fraud’. In other words, the student uses generative AI to rewrite existing material (internet sources, or an assignment from their fellow student) without appropriate referencing.
There are some indicators that a text may be written using generative AI, these include:
- Incorrect, fictitious references
- Overly structured, unnatural feeling text
- Internal inconsistencies in reasoning
- Factually wrong, but confidently written texts
Presence of these indicators does not prove the use of generative AI, nor does the lack of them disprove it. The most effective indicator will be to know your students and their progress.
When students submit an assignment, you can check for the use of generative AI using Turnitin. Note that the score generated by Turnitin can only be used as an indication that generative AI was used but does not provide certainty.
How does it work?
- Check the AI writing indicator. This is shown in the right-hand column of Turnitin's similarity report.
- If the AI writing indicator shows a positive percentage, click this percentage to open the AI writing report. In the report, the sections that were likely written using generative AI are highlighted. This can give you a better understanding on how students may have used generative AI.
⚠️ Note that EUR does not have agreements with other tools (such as GTPZero, OpenAI Text Classifier, etcetera). Please consult with your Learning & Innovation team before uploading student data to those platforms.
Turnitin's AI Writing Detection is currently somewhat limited in the submissions it can be used for:
- File size must be less than 100 MB.
- File must have at least 150 words of prose text in a long form writing format.
- File must not exceed 15.000 words.
- File must be written in English.
It is also far from a perfect solution. Please take the results with a grain of salt. Remember that:
- Turnitin can only calculate the probability that text was AI-generated. Contrary to traditional plagiarism checks, this is not hard evidence.
- Turnitin suffers from both false positives and false negatives. Additionally, Turnitin claims 98% certainty when marking text as being AI generated. However, this has not been independently verified.
- Turnitin cannot differentiate between different types of uses of generative AI. For instance, if the student used ChatGPT to rewrite their own text, this will likely also be marked as being AI generated text.
- Turnitin can only detect output from ChatGPT. If a student uses another generative AI model, this may not be detected.
Determine the risk of generative AI on the ability to assess your students. Based on the results, you may choose to make changes to the way you assess. If you have questions or need advice, contact your Learning & Innovation team.
🟢 Exams are not affected by generative AI. These are supervised in a controlled environment, and students cannot use these tools. There is no need for further action.
🟠 Other types of assignments are potentially affected by generative AI. Proceed to step 2.
Try to complete your assignment using ChatGPT (for writing well-structured texts) and Bing Chat (for information about the content).
Remember that the output you obtain is only one possible outcome. Depending on the use of these tools, other results are also possible. Really take some time to practice with generative AI to experience what it can do.
Be specific: The more specific your prompt, the more accurate the response you're likely to get. Instead of asking broad questions like "What is the meaning of life?", try asking specific questions like "What are the major philosophical theories on the meaning of life?"
Provide context: When you provide context for your prompt, generative AI can better understand what you're asking and provide a more accurate response. For example, you could first enter some content from your course, before asking ChatGPT to complete your assignment.
Experiment with different prompts: Try different prompts and see what kind of responses you get. Experiment with different phrasings and wordings to see how ChatGPT responds.
Make the problem smaller: Instead of simply copying your entire assignment, ask ChatGPT to complete a specific part of your assignment
Provide feedback: If ChatGPT doesn't provide the answer you were looking for, provide feedback on the response so that the model can learn and improve over time.
Difference between ChatGPT and Bing Chat: Bing Chat provides more accurate answers. However, it is tuned to provide brief answers. ChatGPT is better suited for (re)writing longer texts, but lacks some of Bing's accuracy.
- How does the generated output relate to your learning objectives of your course?
- Does the level of the output match the intended level of your learning objectives?
Mainly focus on the content of the output produced by generative AI.
Indications
🟢 The level of the output is insufficient, or it is unusable for your assignment.
🟠 The output is usable and helpful but not fully at the level of the learning objective.
🔴 The output is usable and at the level of (or exceeds) the learning objective.
- Is it still possible to assess the student’s own contribution using the assessment criteria?
- Will you be able to differentiate between students simply using generative AI and students adding their own insights?
- Using your assessment criteria, can you still check that the student has met the learning objective?
Indications
🟢 Using the criteria, the ChatGPT output is essentially irrelevant. You can still fully assess the student’s own contribution.
🔴 Using (all or some of) the criteria, it is not possible to assess the student’s own contribution.
If you assess writing skills, be aware that generative AI can always be used to:
- Develop an outline for the assignment and help students to structure texts.
- Paraphrase literature and other sources.
- Improve quality of writing (spelling, grammar, flow) and tone of voice.
You will need to communicate to students if and what use of generative AI is allowed for your course/assignment. You can follow the steps below to determine what use fits with your assignment and learning objectives.
Before you get started, remember that some use of generative AI cannot be ruled out. Prohibiting it is unlikely to be an effective approach. Firstly, enforcing such policy for non-supervised assignments is difficult, if not impossible. Secondly, in the AI assessment guideline you mitigated the impact of generative AI.
Instead, using generative AI as support (writing, brainstorming, etc.) should be fine for most courses. Still, clearly explain what is accepted and what is not. Moreover, you could advise against (as opposed to prohibiting) using generative AI for (parts of) your assignment.
This step may be most vital to get the desired result. Clearly explain and discuss with students why you recommend a certain use of generative AI.
In particular if you recommend students not to use generative AI, make sure they also understand why they benefit from doing the assignment without it. What do they get out of the process?
You may want students to clearly indicate if and when they used generative AI. Some options include:
- Referencing the use of generative AI
- Report the use of generative AI in a separate section of the assignment. The student should explain how they used the tool.
- The previous option can be extended to include some reflection, for instance: why did the student use generative AI? Was the result useful?
Articles
- Community for Learning & InnovationContact the CLI if you would like to share your expertise and experiences around AI, and if you have any questions or additions to this page.Email address
- Learning & Innovation teamContact your faculty's Learning & Innovation team if you have specific questions about using AI within your teaching or faculty.
Internal resources
The information on this page is collected by the EUR task force AI. Last update: 17 January 2024.