Step 4

Work the Process and Understand Learning

AI for Academic Excellence Student Toolkit

Rule of thumb: Let the workflow guide your process, keep your judgment in the lead, and keep a record of your work.

Writing an essay, solving a case, coding, and preparing a presentation each call for a different approach, and the way you use AI should reflect those differences. The Guides by Assignment Type page gives example workflows.


Keep a Record of Your Process

Academic integrity at UVA rests on disclosure and on the evidence in your process. The University discourages instructors from using AI detectors because they are not reliable, and it asks them to look at process evidence such as notes, drafts, reflections, and check-ins instead. See the FAQ “Should I use AI-detection tools?” Keep your notes, drafts, version history, and important prompts and outputs (Monash University). A record lets you show your thinking if questions arise, and looking back at it is a good way to see what you learned.


AI Across Your Learning Journey

This progression comes courtesy of Jill Mitchell. In every field, learning is a spiral. You revisit skills at deeper levels, and AI’s role changes as you do.

Learning levels and AI's assistance role and UVA examples.

Stage

AI's Role

Examples from Across UVA

Beginning learner

Tutor. AI explains concepts and gives you practice.

Clarifying the difference between mitosis and meiosis in biology, or between assets and liabilities in accounting.

Developing learner

Practice partner. AI tests your understanding and gives feedback.

Quizzing yourself on medication dosage calculations in nursing, or on free-body diagrams in engineering statics.

Advanced learner

Collaborator. You and AI generate and refine work, moving from practice to production.

Drafting a data-cleaning script for a public policy project and then testing it yourself, or building a flexible budget scenario in accounting.

Expert

Co-creator. You combine your expertise with AI to push boundaries.

Using AI to search a large archive of historical letters for patterns and then framing your own research question, or analyzing a 10-K to spot unusual ratios and design an audit strategy.

Some tasks AI already handles well, such as summarizing a standard reading, writing boilerplate code, or preparing basic financial statements. Spend your time on the skills that remain yours: judgment, ethics, creativity, and strategy.

This progression describes how well you know your subject. The AI literacy levels in the AI Literacy Self-Check describe how well you know AI. A student can be an expert in their field and still be building basic AI literacy.


AI Agents

Getting limited help from a chatbot is different from handing an entire task or workflow to an agent that acts on its own. An agent that runs the whole workflow has not left you a process you worked. ChatGPT’s agent mode, which can browse websites and take actions for you, is one example. Ask your instructor before using an agent on coursework. See the FAQ “May I use an AI agent to complete coursework?”

If you do use an agent, these rules apply:

  • Never give an agent, browser extension, or online service your NetBadge password, a logged-in session, or any other University credential. University policy IRM-002 prohibits sharing credentials, and that covers software as well as people (UVA FAQs).
  • Do not run an agentic browser while logged in to Canvas, SIS, Workday, or other University systems. An agent with your login can read records, send messages in your name, and post assignments, and its actions are treated as yours.
  • Use only agents and connectors that ITS has approved, and give the agent the least access the task needs. Choose read-only access unless it needs to change something (Stanford University IT).
  • Confirm each external action yourself, such as sending, posting, deleting, or changing permissions, and don’t leave an agent running unattended.
  • Know what tools the agent can use, what information it can see and keep, and what permissions you are granting. You remain responsible for what it does.
  • Treat text the agent reads, such as web pages, emails, or documents, as possibly containing hidden instructions (see “Put pasted text in a box” in Step 5).

Good practice for using agents to support learning is still forming. Some students are building their own study tools with agents. If you experiment, check with your instructor first and keep the thinking the course asks of you in your own hands. For a deeper treatment of how to give agents the context they need, see Lo’s PASTAS checklist (Lo, 2026b).

Coding Agents

Coding agents such as Claude Code and OpenAI Codex can read, write, and run code across your files. Many students meet them in internships. UVA has not licensed dedicated coding agents for students, so University data protections cover them only if they appear on the ITS list. If your course permits them:

  • Review, understand, and test every change the agent makes (Northwestern University Information Technology).
  • Know which files and data the agent can access or modify.
  • Keep passwords, API keys, and other credentials out of the agent’s reach.
  • Use version control and backups so you can review and reverse changes.
  • Check copyright and licensing for code the agent produces or pulls in, and record the tool and model version you used.

Theses, Dissertations, Exams, and Research

Graduate and professional programs, and undergraduate research, bring rules of their own.

  • Qualifying and comprehensive exams. Ask your program before using AI to prepare for or complete them, and get the answer in writing.
  • Theses and dissertations. Agree with your advisor and committee on how you will use AI, keep a record, and disclose it the way your school requires.
  • Research with people or sensitive data. AI use falls under your IRB protocol, data security requirements, and sponsor terms. Keep research data and unpublished manuscripts out of tools that aren’t approved for them.
  • Publishing. Journals and publishers set their own rules. Journals that follow COPE guidelines do not accept AI tools as authors and expect you to disclose AI use (COPE, 2023). The UVA Library’s Publishing with AI guide collects publisher policies.
  • Peer review. Don’t upload a manuscript or proposal you are reviewing to an AI tool unless the journal or funder allows it.

The sources cited on this page are listed in the Toolkit References.