The temptation for educators is understandable: implement “AI-proof” practices like handwritten assignments and oral assessments, or enter an arms race with AI plagiarism detection tools. However, as AI capabilities rapidly evolve, keeping pace becomes impossible. Faculty find themselves burdened not only with designing new assignment types but also with the challenging task of proving plagiarism – a process that, if mishandled, can damage community relationships and trust.

We found ourselves constantly asking:
How can we support students and families as they navigate these uncharted waters?

A Relationship-First Response

At SBS, our approach to integrating AI into student formation centers on championing relationships with students as image-bearers of God. Like many institutions, we established an AI Task Force and developed a comprehensive AI Policy, complete with traffic light imagery posted in every classroom. We refined our academic integrity standards around three pillars: Originality, Source Citation, and Transparency. Faculty and student assemblies reinforced these expectations within our school’s mission framework.

Yet we soon realized these measures weren’t sufficient. Students and teachers needed more concrete support navigating the academic integrity process. This recognition led us to develop a structured AI Checklist for suspected plagiarism cases.

The AI Checklist Framework

When a teacher suspects inappropriate AI use, they report it to the Dean of Students office. Our deans then apply a five-point checklist, proceeding with disciplinary action only when at least three items are conclusively met. This approach distributes the burden of determination across multiple criteria and personnel, acknowledging the severe limitations of AI detection tools while allowing students to demonstrate their authentic voice and process.

The checklist emphasizes relational knowledge. Faculty are expected to know, care for, and understand their students – a principle woven throughout our evaluation process. Teachers should recognize their students’ typical writing voice and style, and students should have a chance to speak into the process. For this reason, no single point can result in a “guilty” determination and a critical mass of three points must be reached to proceed in the disciplinary process. 

The AI Checklist

1. The paper fails two online plagiarism checkers for AI.
   1. Explanation → SBS uses TurnItIn and GPTZero. The assignment must be identified as AI by both of these.

2. The paper has no version history or checkpoints.
   1. Explanation → The student cannot demonstrate the progress and evolution of their writing over the assignment. This is especially important for long-term assignments. Very few edits, or a rapid appearance of large chunks of text, will not be considered a proper version history.
   2. Students are encouraged to use Google Docs, Microsoft Word, or other word processing software that uses versioning to automatically save their draft history.

3. The student cannot orally defend or explain their writing’s topic, argument, or diction.
   1. Explanation → The student cannot summarize their essay’s argument or exhibit a reasonable comprehension of the contents of the essay. This is not meant to be a trial, but should be an informal dialogue about the essay. If the student cannot discuss the essay’s contents in appropriate depth, then it may be a sign that they were not involved in the process of writing it.

4. The writing sample does not favorably compare to a portfolio of the student’s previous work.
   1. Explanation → Based on a comparison by the teacher(s) and deans to other work written by the student, the writing sample does not contain the student’s “voice” or share a similar writing style and level of depth.

5. The writing sample contains multiple characteristics consistent with AI-generated writing.
   1. Explanation → In the opinion of the teacher(s) and deans, the writing sample contains some or all of the following characteristics:
      • Excellent grammar and syntax, including high use of punctuation
      • Unusual or overly complicated sentence construction
      • Monotonous and impersonal tone and style
      • Consistent use of filler words and/or subordinating conjunctions
      • Usage of notation or vocabulary not typically used in class

Early Results and Reflections

How has it worked? Pretty well!

By the time cases reach parents, we can demonstrate our collaborative approach rather than hasty judgment. Parents appreciate seeing how students can showcase their creative process and how multiple adults ensure fair evaluation. Faculty value the distributed responsibility for reporting, while administrators can better support both teachers and students through clear expectations.

Certainly, the checklist and process require continued refinement. However, we believe this approach creates meaningful opportunities to partner with students and families, clarify expectations and policies, and support all stakeholders in the disciplinary process.

As educational institutions continue adapting to AI’s presence, prioritizing relationships and collaborative assessment may prove more sustainable than technological arms races. Our experience suggests that when students feel known and supported, academic integrity becomes less about detection and more about formation – exactly what education should be.