A Mosaic Approach to Academic Integrity in the AI Era (With Chris Ostro)

Generative AI has developed a dominant presence in educational discourse. Professor Christopher Ostro provides a “Mosaic Approach” to help educators navigate our current AI landscape.

Students’ AI use became apparent when they turned in assignments with content that did not align with the assignment’s word count, formatting restrictions, and number of sources required. These submissions did not work within the scope of the class and were often turned in fully without outlining the process. A large majority of his students using AI referenced the same handful of sources, with some being hallucinated.

What Doesn’t Work

Banning all AI use as a response was not a productive response. This did not account for the ethical ways students may use AI to improve upon existing work. Constantly working to catch students in the act of AI use and seeing it all as cheating damaged student relationships. Running each prompt through ChatGPT and haphazardly using AI detectors, hoping to get similar results as students, was time-consuming. Teaching became unenjoyable, as well.

Since most students are already using AI and want to understand it, Professor Ostro explains that his strict policy created an abstinence-only environment. Creating clear guidelines and training students on proper AI-usage is a more realistic approach.

It is also important to understand where students are coming from. In a worsening job market and increased cost of living, students are reportedly having to work more hours while keeping up with their education. Some students who were at critical learning stages during COVID are missing key skills. They are also having to navigate courses with varying AI policies.

What Works (For The Most Part): Mosaic Approach

Mosaic Approach

AI Literacy Assignments: Students need guidance on how to use AI. The majority of them are using it without proper media literacy and skepticism of the ways AI can be wrong. As an instructor, you can’t teach this unless you are familiar with these tools.

AI Disclosure Form: This gives students the opportunity to be honest with their instructors to prevent Honor Code violations. It’s imperfect because some may forget or lie, but it is important to emphasize its transparency. Putting this form in announcements, on every prompt, on the syllabus and having its own page with video/text makes it harder for students to forget. Here is an example.

AI Detection (with transparency): AI detection has gotten better with lower false positive rates and are more reliable than the human eye. These tools are still imperfect and require human discernment. A high Ai score might be indicative of ethical use rather than serious academic dishonesty. FERPA compliance is important, so be sure to confirm with your school’s IT department on the tool's privacy policies before use.

Process Tracking: GoogleDocs and Office365 have detailed version histories. If an instructor notices the use of tools that fake a document’s version history, there’s no doubt the student is acting in bad faith. The version history may be off because of travel or bad Wifi. A student may also not want to share version history for reasons like privacy concerns.

Awkward Conversations: From Professor Ostro’s anecdotal evidence, most students are not experienced liars. When met with understanding, they are more likely to be honest. Bringing evidence and pointing to previous instances of academic dishonesty helps the conversation.

Second chances: Research by Dench & Joyce shows that students are less likely to cheat again once caught. Professor Ostro notes that giving second chances “allows you to act with much more certainty, encouraging the students exploring this new tool (perhaps clumsily) while holding bad faith students accountable”.

Pros: The Mosaic Approach has allowed for more honest communication with students. Fewer students needed to be Honor Coded while more students corrected mistakes and self-reported. When students were identified for Honor Code violations, this was 100% accurate. Most feedback has become about helping students understand how to improve. Teaching is back to being fun!

Cons: While the results improved academic honesty, there is still some uncertainty about dishonest students. The Mosaic approach requires faculty to be up-to-date with current technology, which is a significant time commitment. Professor Ostro is not convinced of its scalability for faculty who are unfamiliar with AI, students who are more cynical/degree-oriented, and for asynchronous courses.

Contact Information

If you have any questions, feel free to reach out to us at Pangram. Professor Ostro’s contact information is below:

Email: Christopher.Ostro@colorado.edu

LinkedIn: https://www.linkedin.com/in/ochristo/

BlueSky: https://ochristo.bsky.social/