AI-proofing Assignments: Alternative Approaches to AI Detection

With the rise of AI use many questions have been raised related to cheating prevention in the AI era. When AI is capable of so many amazing things, what are the best strategies for protecting assessments from AI? Many educators are seeking AI-proof assessment strategies to ensure student engagement and competency. There are many approaches to AI policies and deciding the permitted, acceptable, and responsible use of AI by students and the identification and detection of AI in student work. 

Recently, a significant amount of research has focused on the efficacy of the various AI detection products that are available and the potential of the underlying science of AI text identification using statistics, machine learning, and artificial intelligence. While AI detectors may be used for information, many educators question their use for discipline or investigations and may constitute an unethical use of AI in education1.

As you navigate assessment in the classroom here are alternative methods for protecting assessments from AI, ensuring active learning, participation and authentic contributions to the goals of your classroom by replacing out of class written artifacts as end products.

Cursive’s browser extension and LMS plugin provide complete transparency for the writing process.

Assessment

‘Viva’ (oral defense and assessment)

To be able to speak to a topic and answer questions about it, a student has to have a deep understanding of the materials and their thinking on a subject2

Education, specifically higher education, has a long track record of using oral defense and Q&A sessions with students to verify their understanding of a concept or providing deeper analysis and thoughts on a written submission they might have made. Many presentations already include this as part of the Q&A session helping evaluators measure understanding. 

Video-based submissions 

Similar to a viva, a video based submission might help to reduce student angst while providing opportunity for them to answer questions or provide a short presentation on the topic. 

Many learning management systems have tools that facilitate the ease of submission though most cell phones have the capability of providing easy video recording that can be done without any additional equipment. Video based submissions formats should be clearly stated to ensure that students are meeting all criteria. 

While AI ‘deepfake’ technology may render this obsolete in the future, the current limitations of this submission process provide students ample opportunity to practice to the format creating a polished performance rather than showcase unscripted creativity. 

Presentations

Group or individual presentations provide students an opportunity to provide a deep dive into a topic, showcase their creativity and critical thinking, while requiring deep understanding similar to a viva. 

Presentations can be a great way to assess multiple formats of communication (writing, speaking, body language) especially beneficial for workplace experience. 

Project and multi-part work

Projects that include multi-modal products: visuals, graphs, charts, spreadsheets, presentations, literature reviews, etc. create a variety of submission requirements that reduce the need for AI detection. 

Simulations and labs

Already employed in many subjects, classrooms, and professions, simulations offer a great opportunity for students to engage with replicas of practical application of their skills. The very best may include opportunities for VR/AR peripherals, while others can be easily completed with a computer or even game materials in a face to face environment (dice, paper, cards). Simulations are harder to design, but also harder to “outsource” to AI reducing the need for AI detection. 

In-class written tests

One of the simplest alternatives is to move writing for class back into the classroom under supervision. This raises the stakes of the submission and severely limits student thinking time to a more constrained 50-90 minute session depending on the class length. 

This can help to invigilate summative assessments more consistently, reducing the need for AI detection but could have adverse impact on students who struggle with handwriting, typing, or simply would benefit from freedom to choose the time and place for writing. 

Proctoring

Proctored examinations, whether face to face or web-based, can be used as alternatives to AI detection tools due to the invigilated nature of their delivery. Proctored exams allow for a wider variety of assessment techniques and options (multiple choice, essay, short answer, etc.) potentially creating more efficient assessment while maintaining high integrity. 

Many web-based proctoring solutions are sufficiently advanced to include anomaly detection, biometric verification, and even live personnel working with the student. Paired with tools like secure exam browser and other secure exam technologies can help. However, many have raised privacy and AI-related concerns about bias, fairness, and invasiveness of AI surveillance methods. 

Policy

Policy and course structure can also help with AI-proofing a class, depending on your circumstances you may be able to modify day-to-day course sequence and structure to build a culture of authenticity without changing assessments. 

Conferences and Check-ins

Many educators will agree that the best way to AI-proof your assessments is to know your students. AI-based writing often includes some specific “tells” making it easy to identify, but with frequent check-ins with students you can be privy to their progress and creativity from the start. Opportunities or requirements for meeting with you to discuss assignments, their strategies for completing it, their research process, and working through early drafts and notes are all ways to ensure authentic assessment submissions. Cursive’s browser extension can easily facilitate the sharing of early information, from the very first bullets of brainstorming to the final round of edits, to help guide conversations toward the best possible final submission. 

Clear AI Assessment Scale3

Clarity in assignment submissions and academic integrity starts in the syllabus, and is reiterated on every assignment posting. Letting your students know exactly the expectations of AI-use, and showing them proper examples is a great way to build trust and dispel the “magic” of AI generated content. 

Revision history & proof of effort

Lastly, revision history for written assessments is a great way to see how the writing process progressed. Google Docs, Track Changes in MS Word, and other tools can provide history of the writing process. Cursive’s browser extension was specifically made to capture, analyze and share this information providing teachers and students both a proof of effort, summary statistics, and revision history for writing. 

  1. https://www.academicintegrity.eu/wp/ ↩︎
  2. https://www.dcu.ie/teu/academic-integrity-principles ↩︎
  3. https://open-publishing.org/journals/index.php/jutlp/article/view/810/769 ↩︎