AI Detection in Student Assignments: How Graidable Helps Norwegian Universities

With the emergence of AI tools like ChatGPT, Claude, and Gemini, Norwegian universities have faced new challenges in ensuring academic integrity. Many institutions have developed guidelines for the use of artificial intelligence, but how can teachers and institutions actually detect when students have used AI to write assignments?

The Challenge of AI in Academic Work

Norwegian universities have quickly adapted to the new reality. As NTNU writes in its guidelines: "It is permitted to use AI as a tool in the preparation of student assignments. If you use AI tools, you can account for your use of the tool either in the assignment or in a separate declaration."

But the guidelines also set clear requirements. Høgskolen i Molde emphasizes: "The most important thing is that texts submitted for assessment, exams, and similar must be written by you." The University of Oslo (UiO) is even more explicit: "All aids are permitted, but it is not permitted to generate entire or parts of exam answers using AI tools such as Chat-GPT."

Guidelines from Norwegian Universities

NTNU – AI Declaration and Responsible Use

NTNU has developed comprehensive guidelines for the use of artificial intelligence. The institution permits the use of AI as a tool but requires students to declare its use. This presents teachers with the challenge: How to verify that the declaration is correct, and how to detect undeclared use?

UiT The Arctic University of Norway – New Guidelines from October 2024

UiT published new guidelines for the use of AI in exams in October 2024. These guidelines reflect the rapidly evolving situation in AI technology and emphasize the need for tools that can help teachers identify unauthorized AI use.

University of Oslo – Clear Boundaries for Exams

UiO has clear guidelines that distinguish between permitted and non-permitted use of AI. For exams, the rules are clear: AI cannot be used to generate answers. This requires teachers to have tools to identify when such boundaries have been crossed.

Høgskolen i Molde – Focus on Original Work

Høgskolen i Molde emphasizes that work must be the student's own. This requires methods to verify that texts are actually written by the student, not generated by AI.

How Does AI Detection Work?

AI detection is a complex task that requires advanced algorithms. Traditional methods such as searching for specific patterns or phrases are no longer sufficient, as modern AI models can produce text that is very similar to human-written content.

Technical Approaches

Modern AI detection systems use several techniques:

1. Stylistic Analysis: Examines writing style, word choice, and sentence structure to identify patterns typical of AI-generated text
2. Semantic Analysis: Assesses the complexity and coherence of content
3. Statistical Modeling: Uses machine learning to identify characteristics that distinguish AI-generated text from human-written text
4. Contextual Analysis: Evaluates whether the text fits into the student's previous work and knowledge level

Challenges and Limitations

It is important to understand that no AI detection system is 100% accurate. As several experts have pointed out, it is impossible to guarantee perfect detection due to the complexity of both human and AI-generated text. Therefore, it is important to use detection as a tool to support the teacher's professional judgment, not as a final verdict.

Graidable's Approach to AI Detection

Graidable has developed an advanced AI detection function that is specifically tailored to the needs of Norwegian universities. Our system combines multiple detection methods to give teachers a thorough assessment of the probability of AI use.

Integrated into the Assessment Process

Graidable's AI detection is integrated directly into the assessment process. When an assignment is uploaded, the system automatically analyzes the text and gives the teacher an indicator of the probability of AI use. This makes it easy to identify assignments that require closer review.

Support for Norwegian University Guidelines

Our system is designed to support the specific guidelines that Norwegian universities have developed:

• NTNU: Graidable can help verify AI declarations by comparing detection results with the student's statement
• UiT: The system supports teachers in following the new guidelines from October 2024
• UiO: Graidable helps identify potential violations of exam rules
• Høgskolen i Molde: Our tools support verification that work is the student's own

Transparent Reporting

Graidable provides detailed reports that show:

• Probability score for AI use
• Specific indicators that influenced the assessment
• Comparison with the student's previous work
• Recommendations for further follow-up

This gives teachers the information they need to make informed decisions, in line with the institution's guidelines.

Best Practices for Institutions

Based on guidelines from Norwegian universities, we recommend the following approach:

1. Combine Detection with Dialogue

AI detection should not be the only factor in an assessment. Combine technical analysis with dialogue with the student. This aligns with NTNU's approach to responsible AI use.

2. Use Detection as a Tool, Not a Judge

As several experts have pointed out, it is impossible to achieve 100% accuracy in AI detection. Use Graidable's tools as a support system for your professional judgment, not as a final decision.

3. Be Transparent with Students

Inform students that AI detection is being used, in line with principles of transparency and fairness. This can also serve as a preventive measure.

4. Follow Institutional Guidelines

Each institution has its own guidelines. Graidable can be customized to support the specific requirements of your institution, whether it's NTNU, UiT, UiO, Høgskolen i Molde, or other Norwegian universities.

5. Document the Process

Use Graidable's reporting to document the assessment process. This is important both to protect the teacher's decisions and to give students the opportunity to appeal.

Ethical Considerations

It is important to emphasize ethical aspects of AI detection:

• Fairness: All students should be treated equally, and detection systems should not discriminate
• Transparency: Students should be informed about the use of detection tools
• Proportionality: The response to detected AI use should be proportional to the severity
• Right to Appeal: Students should have the opportunity to explain themselves and appeal decisions

Graidable is designed with these ethical principles in mind, and our tools give teachers the opportunity to make informed, fair decisions.

Conclusion

Norwegian universities have taken a proactive approach by developing guidelines for the use of artificial intelligence. NTNU, UiT, UiO, Høgskolen i Molde, and other institutions have all established clear frameworks for what is permitted and not permitted.

Graidable's AI detection features support teachers and institutions in following these guidelines by providing tools to identify potentially unauthorized AI use. But it is important to remember that technical detection can never replace the teacher's professional judgment and dialogue with students.

By combining advanced technology with human judgment, we can ensure that academic integrity is maintained while adapting to the new reality of artificial intelligence in education.

Get Started with Graidable

Are you interested in learning more about how Graidable can help your institution with AI detection? Contact us for a demo or learn more about our features on our website.

Graidable supports Norwegian universities in their work to maintain academic integrity in the AI era. With our advanced detection tools and adaptation to Norwegian guidelines, we can help your institution navigate the challenges of artificial intelligence in education.