AI Similarity "Detection"
There are good reasons you'd want your students to not use GenAI and holding students accountable for doing the work and learning is a necessary part of our classes. AI similarity "detection" software is one tool that may indicate GenAI usage. AI similarity "detection" software tools suggest a likelihood that a product same structure that a specific LLM would have produced. Some only "detect" ChatGPT output whereas others can check across LLMs. Unfortunately, there's "beat detection" software that tell students what/where/how to change GenAI output for it to be undetectable. AI similarity "detectors" results can be used in the process of determing academic misconduct but are not proof positive that students used AI to complete the assignment. AI similarity "detection" software programs are subject to bias and inaccuracy*. We recommend if you use them, make sure your students know you are using AI Detectors (as you are inputting their work into AI), use with caution, and that it is not used as an alternative to having a conversation with students.
More successful approaches to reducing unauthorized use includes proactively reflecting & redesigning elements of your curriculum. You can review strategies that discourage GenAI use/help prevent unexpected GenAI use and consider implementing a few. This includes establishing and engaging students with your GenAI policy.
*Here are some references about AI Detection Software
- Elkhatat et al. (2023) Evaluating the efficacy of AI content detection tools in differentiating between human and AI-generated text. International Journal for Educational Integrity.
- Coffey, L. (2024). Professors Cautious of Tools to Detect AI-Generated Writing. Inside Higher Ed
- Liang et al. (2023). GPT detectors are biased against non-native English writers. Patterns
