Guidelines for Integrating Generative AI in Classroom

Bielewicz, Brady; Cynthia Chew; Heeyoung Lee; Morris, Jonna; Wilkenfeld, Daniel
(alphabetic order)
Version: February 16, 2026

Purpose and Scope

These guidelines provide a practical framework for responsible, learning-centered use of generative AI (GenAI) tools in teaching and assessment.

Overall Guideline: Privacy, Confidentiality, and Intellectual Property

Do not upload unpublished research data, patient records, student work (yours or someone else’s), or any other sensitive or identifiable personal information into public GenAI tools (e.g., free consumer versions). Examples include full names; personal addresses or phone numbers; student IDs; medical record numbers; full-face photos; and quotations from reflective writing or discussion posts that could reasonably identify an individual.

Content entered into public GenAI platforms may be retained and handled under the provider’s policies and settings, and may be used to improve the system. As a result, inputs may not be private or secure and can create risks to confidentiality and intellectual property. Uploading such information may violate HIPAA, FERPA, contractual obligations, or institutional policies.

Defining Acceptable GenAI Use

Instructors should specify one of the following permission levels for each assignment or assessment:

Key Guidelines for Students

• AI is not an author: Students may not list ChatGPT, Gemini, or other GenAI tools as a co-author. GenAI is a tool (similar to a spellchecker or calculator), not a collaborator.
• Citation and disclosure are required when GenAI is permitted: If GenAI is used to generate ideas, language, code, or other content, students must acknowledge the tool following the required citation style (e.g., APA, MLA) and provide any assignment-required disclosure (e.g., an AI Use Statement).
  • APA example (adapt as needed):
    • Reference entry: OpenAI. (2023). ChatGPT (Feb 13 version) [Large language model]. https://chat.openai.com
    • In-text citation: (OpenAI, 2023)
• Verification is the student’s responsibility: Students are fully accountable for the accuracy, originality, and integrity of their work, including any errors or fabricated information (“hallucinations”) produced by GenAI tools.

Key Guidelines for Faculty

• Avoid sole reliance on AI detectors for disciplinary action. AI-detection tools can generate false positives, particularly for non-native English speakers.
• State the GenAI policy in the syllabus and in each assignment prompt. Students should not have to infer expectations; specify the permission level, disclosure requirements, and prohibited inputs.
• Prioritize process evidence when appropriate. When learning objectives emphasize reasoning and synthesis, consider grading drafts, revision memos, source-to-claim mapping, and brief oral defenses rather than only a final product.

Assignment and Assessment Examples

Pitt AI Ad Hoc Committee report, 2024 & https://teaching.pitt.edu/resources/teaching-with-generative-ai/ :
Examples of assignments incorporating GenAI:

• Analyzing AI-generated output (e.g., identifying themes or deficiencies).
• Revising an AI-generated first draft (with documented verification and revisions).
• Generating practice questions and self-tests.
• Ask students to complete a written assignment, then use AI to generate a version of the same assignment. Instruct students to compare the two and reflect on their work.
• Instruct students to assign an AI tool a specific persona and roleplay a scenario.
• Have students use AI to make a creative work that helps clarify or illustrate a course concept.
• Ask students to fact check and critique AI output.
• Encourage students to treat AI like a study buddy. Students can quiz themselves on course concepts using AI. Language-learning students can practice their language skills by chatting with AI.
• Teach students prompt engineering and ask them to complete authentic tasks that they will perform in their future professions using AI.
• Give students the option of using AI tools to revise their writing or code.
• Encourage students to use AI to brainstorm and refine topic and research question ideas.

Examples of designs that make misconduct using GenAI more difficult:

• Oral components (brief defenses, interviews, or check-ins)
• Mind mapping (students generate concept maps of arguments they are reading or presenting)
• Process portfolios (outline → draft → revision memo → final)

Uses for AI in Teaching: Course and Curriculum Design

In many contexts, especially large or asynchronous courses, long take-home essays may be less effective for evaluating student learning. Consider assessment strategies that align with learning objectives, class size, and modality.

Assessment Approaches by Class Context

Large classes:

• In-class short writing with a secure platform (e.g., ExamSoft with LockDown): low-stakes in-class assignments and timed reflection prompts (2–3 sentences).
• Frequent formative checks (e.g., Top Hat) to assess participation, comprehension, and application of key concepts.
• Structured critical thinking practice through peer evaluation using clear criteria.
• Proctored testing, group projects that require observable participation (i.e., in-class work session or brief oral check-ins).

Small classes:

• Writing-intensive instruction and assessment.
• Iterative writing assignments with draft cycles and feedback focused on reasoning and evidence use (not only polish).
• Oral defense.

Asynchronous/online courses:

• Synchronous quizzes or exams using institutionally approved secure tools when high-stakes assessment is needed.
• Project- and portfolio-based assessment emphasizing process evidence (drafts, revision memos, source mapping) when secure testing is not feasible.
• Oral check-ins or short tests (live or recorded) to confirm understanding of key decisions.

Evaluating Online Learners With and Without Secure/Identity Controls

With secure/identity (SI) controls (e.g., proctoring or lockdown browsers):

• Timed assessments emphasizing applied reasoning (case-based questions, short answers).
• Short, locked-down writing prompts to sample authentic reasoning.
• Observed demonstrations and brief oral defenses for skills-based courses.

Without SI controls:

• Process portfolios with required artifacts (outline, drafts, revision memo, claim-to-source map).
• Authentic, course-specific prompts that require use of assigned materials.
• Peer review with structured criteria and grading of feedback quality.
• Random or scheduled short oral check-ins to confirm understanding.

AI-Assisted Grading: When Acceptable and Under What Conditions

AI-assisted grading may be used only when all of the following conditions are met:

• The platform is institutionally approved and complies with privacy and security requirements.
• Identifiable student information is not entered into public tools; any data handling follows institutional policy.
• A human reviews and finalizes all grades (AI supports, but does not replace the instructor’s judgment).
• Rubric alignment is preserved and communicated to students.

Uses for AI in Teaching

Resource: https://teaching.pitt.edu/resources/teaching-with-generative-ai/

• Create assignment instructions, grading criteria, or tools like rubrics
• Generate FAQs with explanations for confusing concepts
• Create case studies, scenarios, or examples to respond to or critique in class
• Prepare lesson plans or outlines
• Suggest active-learning activities
• Generate or revise quiz questions
• Create study guides or study games to help students prepare for exams
• Compose comment bank items or sample responses to provide students with assignment feedback
• Respond to student emails about basic questions