AY 128 UC Berkeley Astronomy Data Lab Site logo

Ay128/256 Fall 2025 - Use Policy of Artificial Intelligence (AI)

Ay128/256 Fall 2025 - Use Policy of Artificial Intelligence (AI)

Based on https://github.com/profjsb/ucbastro-ai-policy/blob/main/policy.md

Preamble

The goals of a higher education are often aided by the use of modern software. Artificial Intelligence (AI) can be a powerful tool for learning. However, misuse of AI can run afoul of academic honor codes. This policy document is meant to clarify some of the finer points around the use of AI.

Policy

  • Fair Use: we expect students to use a variety of tools and applications to improve understanding and accelerate learning. AI tools should be used in alignment with these goals, not to circumvent them. Subcontracting academic work to a crowdsourcing platform, a tutor, or roommate would constitute academic misconduct; so too would employing AI in a similar way. When in doubt, err on the side of less usage.

  • Context and Clarity: instructors can explicitly require, suggest, or forbid the use of certain classes of AI tools for certain assignments. When in doubt, students should ask instructors for guidelines.

  • Equitable Use: not all AI tools will be available to all students in the same cohort; and many of those tools could be prohibitively expensive to use. This may lead to classroom inequities. Students should generally only consider using tools that are widely available. If assignments warrant, instructors may choose to provide access to costly tools to all students in a class.

  • Disclosure. The use of AI tools shall be disclosed on all class work presented for formal evaluation (e.g., homework, lab reports). Such a disclosure must provide enough information for a reader/grader to understand the extent to which AI tools were used. For example, at the end of a problem set a student could write “Tool X was used in question 3 to help formulate the complex SQL query” or “No AI tools were used.”

  • Responsible ownership: Any output (direct or indirect) from allowed AI tools for classroom work is considered to be fully vetted by the student. That is, the student is entirely responsible for the correctness (or not) of fully or partially generated AI content in their work.

  • Non-classroom academic works presented as original (e.g., academic journals, thesis chapters, talk presentations) are bound by this policy and the policies of the content publisher (e.g., journal or conference organization) where the work is presented. To the extent that the policies conflict on the matter of AI usage, the more restrictive policy should apply.

Examples

The follow are example usages of AI that augment but do not supercede those given in the Definitions & Examples of Academic Misconduct, including forms of cheating, Plagiarism, and False Information and Representation, Fabrication or Alteration of Information.

Unacceptable Usage

  • Cheating: AI is prompted to generate a substantial portion of the answer to a homework or test question (either plain text or code). This is similar to ``Allowing others to do an assignment or portion of an assignment for you, including the use of a commercial term-paper service.’’ (cf. cheating)
  • Plagiarism: AI is prompted to generate small portions of answers to homework or test questions, such output is used and not credited properly. If you would ordinarily cite a source (or thank a roommate) for a snippet of code or an interpretation of text, then you should credit the tool used.
  • False Information and Representation, Fabrication or Alteration of Information: Using AI tools to ``fix’’ (ie. alter) data obtained in a lab setting with the intent to mislead.

Acceptable Usage

  • AI tools are used to suggest and improve grammar or spelling of a student’s work. AI-backed search engines are used to help a student clarify a misunderstanding.
  • AI generates draft git commit messages on student’s code used for a class assignment
  • AI bot formulates comments and suggests changes on code pull requests. The comments and suggestions are public and the “authorship” is clear.
  • AI system provides brainstormed answers that a student uses as a starting point to write a long-form answer to a question.
  • AI provides a useful solution to a coding challenge for a homework, that solution is (largely) used, and the use of the AI tool is properly attributed.