Artificial intelligence in teaching

• Artificial intelligence (AI) in this context refers to generative AI, such as language models or image-generation services.
• AI is guided by prompts, i.e. commands in plain language sentences. These prompts may include questions, background information and instructions for the AI that help in getting the desired answer.
• The use of AI is measured by tokens. Tokens are a numeric unit utilized by the language model that can represent a character or string, a word or even a short sentence.
• AI outputs should always be checked, as the response may contain errors, outdated information or be incomplete.

On May 30th, 2024, the University of Helsinki’s Academic Affairs Council published the revised guidelines on the use of AI in teaching (pdf). Please note that they may be further specified in the light of future regulation and technological developments.

University policy on the use of AI and large language models in teaching and learning:

1. As a rule, large language models may be used in teaching and in support of writing. On pedagogical grounds, course teachers can restrict or prohibit the use of large language models on their courses.
2. Teachers and students are responsible for the accuracy of the language and content of their written work. The use of large language models must be indicated as defined by course teachers and in accordance with the principles of research integrity.
3. In maturity tests, the use of large language models is not permitted.
4. Faculty councils, degree programme steering groups and the Language Centre may draw up supplementary guidelines on the use of artificial intelligence.
5. The use of large language models on courses must be primarily based on the services provided by the University. The accessibility and information security of services must be kept in mind when using large language models. Students cannot be required to use services other than those provided by the University.
6. If students use large language models on courses, components thereof, or in examinations where it has been prohibited in advance, or fail to report the use of large language models as instructed, this constitutes cheating, which will be handled according to the same principles as other cases of cheating.

The guidelines state that course coordinators decide the extent to which AI can be used in their courses. Please note that AI can never be used for maturity tests.

The University of Helsinki has two generative AI services, CurreChat developed at the University and Micorosoft's Copilot with commercial data protection. Copilot is available without restriction to both teachers and students. CurreChat is freely available to staff, while students may use it on courses registered with the service.

CurreChat is more suitable for educational use in terms of both data protection and features, and future development of CurreChat will focus specifically on educational use. For more information on the differences between the services and on security, see the IT HelpDesk guide Generative AI at the University.

Tell your students if you have used AI tools to aid the production of course materials or learning tasks, or if they contain AI-generated components.

Introducing CurreChat on your course

CurreChat is available at curre.helsinki.fi/chat. The service requires login with university credentials.

Teachers can activate the courses they are responsible for directly from the service. Activation creates a course area where enrolled students can use CurreChat for the duration of the course. Conversations students have in CurreChat are not visible to teachers or administrators, nor are they stored in the system.

In the course settings, the teacher can change the course duration and the token limit available to students during the course, add teachers to the course, and create course-specific prompts.

Prompts are instructions for the AI, guiding its behavior during conversations. With course-specific prompts, the teacher can create tasks for students or otherwise direct the use of AI in the course. In CurreChat, prompts can include extensive text-based supplementary materials that the AI can utilize in its responses.
 

Artificial intelligence can enhance and assist the production of learning materials and courses in many ways. In all cases, it is important to check the content carefully. Provide the artificial intelligence tool with as much background information as possible about the scope of the course, the target group, and the topics you want the answers to cover.  For example, you can ask generative artificial intelligence to: 

• Generate ideas for the course structure, lecture content, and learning assignments. You can ask GenAI to evaluate your course plan, for example, from the perspective of constructivist alignment.
• Create a lecture outline or a slide show template on a specific topic. Create assignments on a specific topic for group work, exam questions, discussion topics, or create images to be used as part of the teaching materials.
• Create practical examples of a topic being taught, explaining difficult concepts in more general terms.
• Review your own content and provide further ideas, correct grammatical errors, or translate text from one language to another. 

Generative AI is becoming increasingly integrated into the learning management systems used in university teaching:

• Moodle's question generator can be used to create multiple-choice questions on the course topic directly in the course area's question bank. Teachers can use GenAI in Moodle's text editor to generate text or images directly in Moodle.  
• Thinglink offers the possibility to create both regular and 360° images with the help of artificial intelligence. In addition, it is possible to create interactive scenario tasks using text materials such as scientific articles.
• When producing videos, artificial intelligence can assist with language translation, subtitling, or speech generation from a given text.  

To ensure transparency, students should be informed about the use of artificial intelligence in the production of teaching materials. As an example: "Artificial intelligence has been used to assist in the creation of exam questions, and teachers have checked the accuracy of the questions and answer options." 

Students can use AI in learning assignments in different ways, depending on the course learning outcomes and the assignment. Visit "AI Assessment Scale (AIAS)" web pages for examples of the roles AI can play in learning assignments and how to instruct students. According to this scale, AI use can be entirely excluded, its use can be permitted in planning and brainstorming, or it can serve as a tool for content production or as an integral part of the learning process without forgetting the critical perspective. 

Tell students:

1. where AI can be used in courses and assignments,
2. what kind of use is prohibited and
3. how to describe and report on AI use.

• Learning assignments and assessment criteria should be clear to students, including in terms of AI use.
• It is also advisable to discuss how the permitted use of AI or any restrictions on it in assignments links to the learning outcomes of the course of assignment. 

Provide instructions for reporting. Consider what kind of reporting is relevant to assessing completed assignments. You can instruct students, for example, in the following way: 

• ‘Describe the AI solution you used, its role, and at what stage of the work you used it.’
• Depending on the assignment type, you can also request the following from students:
  • ‘Indicate (by citation/reference/specific colour) any sections generated by AI in your assignment.’
  • ‘Please include the answer you produced yourself before entering it into the AI solution.’
  • ‘Please include the question (prompt) you submitted to the AI service and the answer you received.’
• If AI is used for completing assignments, it is good practice to add a section for critical reflection that advances students’ AI literacy and skills as well as highlighting the learning process. Critical reflection should focus on both the use of AI in the process and the content to be learned. 
  Examples of reflection questions:
  • ‘Please include a reflection on how the assignment completed using AI changed your work and whether the answers were useful. Limit the reflection to roughly x words.’
  • ‘What information was missing from the AI-generated answers?’
  • ‘What biases can be inferred from the AI-generated answers?’
  • ‘How do the AI-generated answers compare with academic literature and other course material?’
• Examples of instructions:
  • ‘In this assignment, you can use AI to help in brainstorming and language revision, as well as ask feedback for formulating the content of your answer. The submitted assignment must not include any text written by AI solutions. Report on AI use by describing the solution or solutions you used, their purpose and the stage of assignment completion where they were used.’
  • ‘You can use AI freely to help you complete this assignment. However, your submitted assignment must correspond to your personal understanding and knowledge of the topic. The last session will include a group discussion, where you must be able to justify the statements you have written and actively participate from a perspective of your choice. The use of AI must be reported by completing a reflection assignment. The instructions for completing it are included in the assignment itself.’

Determine how AI responds to students’ prompts 

When adding assignments featuring AI use to teaching in a pedagogically appropriate manner, you as the teacher should draw up instructions for the AI solution to guide the discussion between it and students. These "course prompts" provide background information for how the AI solution should respond to students’ questions or prompts. Such instructions can be given in CurreChat's course view or in the mooc.fi service.

With course prompts, you can

• Design learning assignments where AI is a learning and practice partner, including instructing it to ask students questions on course topics, preparing them for the examination
• Personalise teaching by instructing the AI solution to adjust its answers and questions according to individual students’ skills
• Create AI patients or other virtual characters for practical training and simulations
• Create a formative feedback provider who is instructed to examine specific matters in students’ answers

AI solutions can also be provided with background and source material on which to primarily base their answers.

• CurreChat enables teachers to add background material to any framed chat discussions in their courses.
• The source material feature uses what is known as a retrieval-augmented generation (RAG) functionality, which enables AI to be guided to retrieve information and to base its answers on data uploaded to the solution, such as certain articles, writing instructions or course material.
• Any files added to source material should be in text format.

Consider the role of AI assignments as part of courses

Learning assignments and what we share about their assessment markedly affect students’ actions. Consider what kind of learning assignments your courses as a whole include, how different assignments together promote and enable the achievement of learning outcomes, and how as the teacher you can monitor and assess learning. 

• Assignments where students use AI are well suited to flipped classroom instruction. In remote periods, topics can be explored with the help of either learning material or AI solutions. In contact teaching situations, it is easy to jointly discuss AI use and the answers it provides, advance related skills and correct any erroneous notions.
• In addition to remotely completed assignments involving AI use, courses can include further assignments that demonstrate students’ skills without AI and help in assessment. These include group discussions based on independently completed assignments and brief presentations on key topics.
• In process-based learning, teachers can monitor students’ learning through assignments submitted during courses and formative feedback as well as request supplementary answers or clarifications during courses for sections that remain unclear.
• Also consider the role of individual assignments in courses as a whole. Assignments easily completed by AI should not significantly affect the overall course grade.

Using AI in assessment

• Also in the case of services provided by the University, differences in data protection (further details on the Generative AI at the University page) must also be considered, which is why the use of AI in courses must be planned so that students themselves enter their text in CurreChat or Copilot.
• CurreChat enables teachers to use instructions, or ‘background prompts’, according to which AI solutions provide feedback on students’ texts. For example, CurreChat can be asked to pay attention to certain aspects in answers, or to provide students with tips on how to supplement their answer based on the model answer in the background. Students should be reminded that feedback provided by AI may contain errors.

Tips for assessment when AI may have been used for learning assignments:

• Add an assignment to the course that demonstrates students’ understanding of the topic without AI, such as a poster, a mini-exam in the classroom or a group discussion.
• Ask students to report on their AI use and reflect on the answers given by AI. Effective reporting guidelines help teachers understand students’ learning processes during assignments and develop students’ AI literacy.
• Ask students to submit assignments during the course and use them to monitor progress. If a student's contribution appears to have been completed by AI, discuss the matter with them already during the course.
• AI can also help in courses, for example, in information retrieval or as a practice partner alongside separately assessed assignments.
• Consider the weight of AI-assisted assignments in course assessment. 

Suspected cheating or plagiarism

• At the moment, it is impossible to reliably determine whether answers are produced by AI or students. Services that assess whether text is produced by AI are available, but their answers are only estimates, which are affected by, for example, the use of AI for revising the language of the text or translating it. In addition, it must be noted that teachers are not permitted to enter students’ answers to services outside the University, which do not necessarily comply with GDPR.
• If you suspect the use of AI contrary to your instructions, discuss the matter with the student. You should highlight the suspicious sections in the student's answer and, through discussion, assess whether their skills correspond to the submitted assignment.
• Suspected academic misconduct is handled in accordance with University guidelines. The page Cheating and plagiarism states the following:
  • “Students are responsible for any major decisions they make concerning their work, as well as any fraudulent activities.  Thus, even if the teacher has not specifically instructed the student to use the right citation style or explained cheating and plagiarism during their course, the student is still responsible for any cheating in their work.”
  • “Credits must be earned honestly, with the student accurately demonstrating their knowledge.”