Navigating Academic Integrity in the AI Landscape

This CCTL resource provides a summary of current thinking and strategies on designing assignments for the AI landscape.

Moving some student work into the classroom is one way to promote academic integrity. Consult this resource for considerations for integrating more in-class writing into a course.

• In Reflections on Teaching Without AI, a panel from Spring 2026, instructors discuss oral exams, low-stakes quizzing, and extended in-class writing assignments.
• Benjamin Morgan (English) discusses an “opt-in/opt-out” approach to AI use.
• Lisa Rosen (Committee on Education) and Darya Tsymbalyuk (Slavic Studies) discuss creative alternatives to take-home essays. 
• Ahmed Abozaid (Committee on International Relations) discusses group presentations. 
• Sarah Johnson (Law, Letters, and Society) and Mehrnoush Soroush (Middle Eastern Studies), discuss scaffolding writing assignments and other projects. 

• Communicate clearly and specifically when AI tools are and are not allowed, and what uses constitute a violation of academic integrity. The AI Assessment Scale may help to spur thinking.
• Focus on specific potential uses (brainstorming, outlining, drafting, etc.) relevant to your assignments, rather than tools (Claude, Grammarly, etc.). 
• When AI use is permitted, communicate when and how it should be disclosed and correctly attributed.  
• Provide reasoning that explains how the policy supports the learning process, which helps students understand the pedagogical rationale for the policy.
• Clearly state if you plan to use the Pangram AI detection tool provided by the University. 
• Reference the University’s academic integrity policy and, where relevant, that of your school, division, department, or program.
• Indicate the potential consequences of violating the policy (e.g. failing the assignment, reporting to the Dean of Students, etc.).
• Consider how word choice and tone will communicate and foster trust in your students. 
• In addition to including it in the syllabus, discuss your policy and expectations on the first day of class and before key assignments.

Research conducted at the University indicates that other AI detectors are not as effective or reliable (Jabarian & Imas, 2025). Pangram is FERPA-protected and they do not use submitted materials to train their model. Visit the Pangram website to learn about data privacy, how the model is trained, and how to delete scans.

Some instructors may choose not to use AI detection tools out of a concern that they create an atmosphere of surveillance and signal distrust of students, pointing to doubts about the efficacy of detectors, the risk of false positives (i.e., that the detector may incorrectly indicate that text is AI-generated when it is not), and the potential for bias against non-native English writers. These are real concerns, and instructors who elect to use Pangram should address them with students. Assure students that a Pangram score alone will not be used to determine inappropriate AI use, that you will have a conversation with a student to learn more about their learning and work process and will gather more information before drawing any conclusion. Convey that those conversations will be focused on the student’s learning and development. 

State clearly in the syllabus and during class that you will use Pangram to scan their work. This may help avoid a student feeling “tricked” if you refer to a Pangram score in an academic integrity conversation. Transparency also may help the tool have a deterrent effect.

You may not run every assignment through Pangram, but when you do, you will want to weigh whether to run every student’s submission (universal) or only those you suspect of inappropriate AI use (select). On one hand, select use may preserve a more “human” approach and decrease a sense of surveillance. On the other hand, universal use guards against bias or idiosyncratic judgment on the part of the instructor. Note that when you use Pangram through Canvas, all submissions will be scanned by Pangram.

Li et al.’s interviews with UChicago instructors found that Pangram and related tools were most useful when they helped instructors understand how a submission may have been produced, identify concrete points for conversation, or reduce manual checking work. They were least useful when they appeared to make a judgment that instructors saw as pedagogical, contextual, or relational.

A Pangram result, in isolation, is not enough to conclude a student made inappropriate use of AI. Rather, it is one piece of supporting information in identifying possible inappropriate AI use. Other considerations that may inform a determination of whether student work makes inappropriate use of AI include the following. See Li et al. for a deeper discussion.

• Comparison with previous student work.
• Comparison with student’s interactions in class discussions or office hours.
• The work discusses ideas or sources not mentioned in class.
• The work includes hallucinated references.
• The student’s ability to orally explain ideas and their work process in a satisfactory manner.
• Information from process tracking or similar practices, such as version history or apps like Process Feedback.

Focus on dialogue and student learning and development. Think about this conversation as an opportunity to ask questions and learn about the student’s process, rather than an immediate accusation of academic misconduct. For example, you might ask about:

• Approaches to the research, drafting, revision, and editing processes.
• Perceptions of the assignment’s value and difficulty level.
• Perceptions of where their work is strongest and weakest.
• Strategies for finding, reading, and evaluating sources.
• Specific passages, claims, or concepts for the student to explain or elaborate upon.

Hearing the answers to these questions, whether they are substantive and specific, and whether they align with the student’s work (previous and the work in question) may illuminate whether and how the student inappropriately used AI—and they are also useful questions for the student to reflect on their learning and study processes.

Be transparent, specific, and honest about why their work raised suspicion for you. Whether it’s because of hallucinated sources, formulaic or bland language or analysis, or significant style or tone differences from their previous work or work they produced in-class.

Consider the full context and possible mitigating circumstances. Students behave in academically dishonest ways for a variety of reasons (Rettinger & Bertram Gallant, 2025), which may shape what responses are most appropriate for the student’s learning and development.

Have a developmental conversation about why they took a shortcut and how they can better approach the work of the course. If they are struggling with the course content, believe their work is not strong, or are having issues with time management, consider connecting them to appropriate learning resources, such as the Academic Resource Center, College Peer Tutors, or Writing Tutors and Coaches. Offer to meet in office hours to discuss course content, the student’s work in progress, and other forms of support.

Consult with or report the incident to the Office of College Community Standards (undergraduate students) or the Dean of Students for the appropriate Division or School (graduate and professional students). 

OCCS (collegestandards@uchicago.edu) is available to consult and advise on responding to academic integrity concerns at any point in the process. OCCS’s process is educational and aims to help the student learn from the incident, prevent repeat conduct, promote community wellbeing, and uphold the University mission.

Academic misconduct concerns involving College students can be reported using this form.

When OCCS receives a report:

1. OCCS contacts the student.
2. The student has an opportunity to respond.
3. OCCS resolves the case, which may include:
   1. Educational intervention
   2. Referral to additional resources, such as librarians or academic coaching
   3. Informal resolution with no disciplinary sanction
   4. Formal administrative resolution with a disciplinary warning or disciplinary probation
   5. Referral to the College Area Disciplinary Committee (CADC) for further investigation and resolution, which may result in no finding of responsibility, disciplinary warning, disciplinary probation, disciplinary suspension, disciplinary expulsion, and other discretionary sanctions
4. OCCS follows up with the reporter when the case is resolved to let them know the outcome (not the sanction, if there is one).