Tips for teachers on using artificial intelligence in teaching

How to communicate about potential restrictions on a course or learning task involving artificial intelligence 

Inform students about how you want to limit the use of artificial intelligence and why in a timely manner. Consult the course's learning objectives when justifying restrictions. AI can be used in various ways, and in some cases, it may be challenging for the student to understand whether the software they are using is powered by AI in the background. For example, translation and proofreading tools may utilize AI automatically. If you are restricting the usage, consider asking students to describe the different tools and methods they have used in completing the task.

• Communicate early about restrictions: You can describe the possible limitations in the course syllabus before the start of the course. The syllabus can be easily accessed and edited on the homepage of your MyCourses workspace. At the beginning of the course, it is also a good idea to review the courses' rules and remind students to read the syllabus.
• Separate instructions for learning tasks: If you want to restrict or specify the use of AI only for specific learning tasks, it is advisable to include the restrictions in the task instructions themselves.

How can I check if artificial intelligence was used in generating the text? 

Currently, there is no completely reliable way to check if a text has been generated with the assistance of artificial intelligence. Aalto University provides the plagiarism detection tool Turnitin, which includes a tool for evaluating the use of AI in text. It's important to note that this tool provides only an indicative assessment and cannot categorically confirm the use of AI. For more information about the tool, please visit https://www.turnitin.com/products/features/ai-writing-detection/faq. Please note that it is not allowed to submit student responses into systems that have not officially undergone any security checks by Aalto University.

If you suspect that AI has been used in text production contrary to the given instructions, you can assess a few aspects of the text:

• Unusual references in the text: AI can generate fictional references upon request. It does not verify the existence of the source, making non-existent references a possibility.
• Outdated information: large language models might be retrieving information from a database that only goes until a certain year, and does not include the latest facts and literature. Therefore, AI-generated text can refer to debunked or old information as current.
• Writing style: Text produced by AI often tends to be fluent, error-free, and natural-sounding. If the text appears exceptionally flawless or remarkably professional, it may indicate the use of AI. You can test your assignments by asking AI to provide an answer and analyse the style of the text as reference to detect misusage.
• Repetition or nonsensical sentences: AI based on large language models may generate sentences that do not convey meaningful content or repeat the same information by varying the wording across different parts of the text.

• If your conclusion is that the student has deliberately acted against the instructions, follow the guidelines for handling such violations.

Artificial intelligence in a learning task

Currently, Aalto University does not have a widely available AI service based on a large language model that teachers and students can use with their Aalto accounts. Although some students may have personal accounts for external AI services, the creation of personal accounts cannot be compulsory for completing a learning task. In case you want students to use external services it is good to remind them about data privacy, which in general lines means that entering other students' answers or anyone's personal information into external systems is not allowed.

• Integrating AI: Artificial intelligence can be used actively when generating task submission. For example, students can study how to refine text or incorporate AI-generated text as a natural part of text. In this type of task, it is advisable to ask students to describe how they have used AI in generating their answers and clearly indicate the AI-generated content. Another option is to ask student to review answers generated by artificial intelligence to enhance critical reflection skills. It is also wise to design alternative method of completion for those who do not wish to create an account for an external system.
• Citing AI: The teacher can ask the student to reference sources and ensure the accuracy of both the content and sources created by artificial intelligence. To cite ChatGPT in APA style, you can follow the guidelines (How to cite ChatGPT) provided by the American Psychological Association (APA). Here's an example of how you can cite ChatGPT:
  
  OpenAI. (2023). ChatGPT (Mar 14 version) [Large language model].
• Narrow your focus: The teacher can provide a limited set of reference materials or a case example as a starting point for the student to generate their response. If the material is relatively recent and the task is designed in a way that requires the student to justify their solution, the benefit of using artificial intelligence is usually quite small.
• Use non-textual elements: Ask students to interpret a graph or an image. Interpreting visual elements such as graphs or images is often much more challenging for artificial intelligence than text interpretation.
• Adapt your own assignments: Teachers too can benefit from AI to design better assignments. You can use AI as an assistant to ideate or test your assignments. One good rule of thumb is ensuring that your assignment can never be completed to a maximum grade merely by using AI, and asking AI to provide sample answers can be useful in enhancing your activity to include criteria that are harder for that kind of tool to fulfil.

Examinations and other assessment methods 

The development of large language models (such as ChatGPT) has made us consider more carefully how to ensure academic integrity and foster student learning in assignments and exams. The following list summarizes what kind of options there are for examinations. The tips listed in the previous section can also be applied in exam design. 

• Remote online exams: Online exams that can be completed flexibly from any location allow students to use all available resources. Monitoring and supervision can be challenging in this format. This form of examination is suitable for applied tasks and situations where there is no specific need to restrict the use of artificial intelligence. It is recommended to use the similarity checker Turnitin when collecting student works if applicable to the exam format.
• Supervised exams: If the exam requires supervision and restricted use of additional material, teachers can arrange exams on campus or use the electronic Exam service that includes electronic exam rooms that are monitored by cameras. Usage of external systems and resources is limited in the Exam rooms.  
• Oral exam or presentation: Face-to-face assessment, either individually or in a group, allows the teacher to assess learning in a supervised setting using selected resources. In-person examinations allow the teacher to interview or ask clarifying questions to gain a more comprehensive view of student learning. 
• Experimentations in the field/laboratory/workshops: Similar to oral exams, experiments in authentic settings, such as laboratories, workshops and other work-related environments, allow students to demonstrate their knowledge and skills in a more unique manner. Depending on the situation, the use of artificial intelligence can be limited or restricted if needed.  

• See also: Digital learning assessment and online examination | Aalto University 

Read also: Guidance for the use of artificial intelligence in teaching and learning at Aalto University and Tips for using artificial intelligence for students