AI Literacy in the Biomedical Sciences

Many teachers object to the use of AI tools in the classroom, and have turned to using AI detectors on their students' work. However, there is mounting evidence than AI detection tools do not work. AI detection tools are rife with both false positives - falsely claiming a work that is human-written was AI-generated - and false negatives - falsely claiming that AI-generated works were human-written. Both false negatives and false positives can be harmful to students in different ways.

This doesn't mean that there is no way to prevent the use of AI in the classroom! Studies have shown that students with higher AI literacy (i.e., students who understand the tools better) are less likely to use AI tools. In other words, taking the time to teach students about AI can help help prevent its unsanctioned use in your class. If you are not sure where to start, check the Resources & Links tab on this guide!

If you would like to see the difficulty of parsing AI- vs. human-generated text yourself, the website Real or Fake Text lets you guess whether sentences are human- or AI-written. The website Which Face Is Real? has a similar exercise for AI-generated images.

False positives happen when an AI detection tool falsely flags human-written material as being AI-generated. False positives are more likely to affect neurodivergent students or students who speak English as a second language. There is also evidence that some tools' published false positive rates are much lower than their actual rate. Anecdotally, some undergraduate university students have spoken with media outlets about feeling frustrated that their hand-written essays were flagged as being AI-generated - to the point that some have transferred schools or dropped out over it.

Additionally, some students will write a full essay themselves and use an AI tool to clean up the grammar and syntax - AI-detection tools cannot tell the difference between works that are fully AI-generated vs. edited by an AI.

If you are curious to see what it looks like when an AI detection tool flags false positives, you can feed it a text that predates generative AI - such as the US Constitution.

Some students use AI humanizing software to make their AI-generated writing undetectable to AI detection software. AI humanizing software can edit AI-generated writing to make it more similar to human writing, and remove certain "tells" that AI detection software rely on. High quality AI humanizers are paid software, which means that students from a higher socioeconomic status may have more access, meaning they may be more likely to "get away" with using AI more than students from a lower socioeconomic status background.

None of this means that teachers can't prevent or lessen the use of AI in the classroom! There are two steps you can take to reduce AI use in your class:

1. Write a clear, direct AI use policy in your syllabus. If some AI use is allowed but not all, be extremely clear with your students where you do or do not allow AI, or what uses of AI are acceptable or unacceptable. Unclear policies just confuse students. There are many examples of AI syllabus policies floating around online - try this crowdsourced Syllabi Policies for AI Generative Tools document to start. CELT has also put together guidance on Developing Syllabus Statements for AI.
2. Take the time to teach your students AI literacy. Studies have shown that students with higher AI literacy are less likely to use AI tools. You cannot assume that your students have learned about AI in another class, or that they are "digital natives" who don't need to be taught about technology. AI literacy is not a standard part of primary, secondary, or higher education yet. If you are not sure where to start, there is a list of teaching tools in this guide.

As always, if you need help, we have resources! You can contact your library at hhsl@tufts.edu, or the Center for the Enhancement of Learning and Teaching (CELT) at celt@tufts.edu.