Research, Dissertations/Theses, and Graduate Students
Drafting and AI-Use Disclosure
General Principles
Human accountability: For all students and instructors, the person is ultimately responsible for the work they produce, regardless of the tools used to create it. We expect that human judgment is applied to all work, even and especially when AI is used as part of the process. Students and instructors are responsible for verifying AI-generated assumptions, calculations, citations, code, conclusions, data analyses, designs, facts, and writing. Students and instructors should regularly discuss ethical AI use and the integrity of data they access or upload.
Development of judgment and skills: We are not here to help students ingest and regurgitate a fixed set of material. Instead, we strive to empower students to think critically and solve problems so they develop judgment and learning skills they can draw on for the rest of their lives.
Based on these principles, the ideas below provide guidance to faculty and students on two categories defined by the students’ perspective: 1) using AI to learn and 2) learning to use AI. Both categories are critical to the experience of all Cullen College students.
These guidelines supplement, but do not supersede, University of Houston AI policies, which can be found here. Given the rapid changes in AI technology, the guidelines should be revisited at least annually.
Guidance for Instructors
The disruptive and constantly evolving nature of AI technology gives instructors an opportunity to model what lifelong learning looks like. It requires that we learn from and alongside our students, which is a good thing.
This guidance is presented as a list of recommended practices in decreasing order of strength of recommendation. The practices at the top are considered strong recommendations for all instructors of Cullen College of Engineering courses.
Using AI to Learn
- AI-aware pedagogy: Courses should not have pedagogy that pretends AI doesn’t exist or only adds AI-policing to an existing course. These approaches are insufficient to motivate students or prepare them for an AI-augmented world. All instructors are expected to adapt and continue to adapt their courses as technology evolves.
- Syllabus language:
- Syllabi should not include blanket statements like, “No AI can be used for this course.” These statements are too vague to be useful, particularly when AI is embedded in every search engine and most coding and word-processing tools. Instead, the college recommends that instructors provide AI guidance at the assignment level, both in the syllabus and in the assignment instructions. By assignments, we mean exams, quizzes, projects, homework, and other graded coursework.
- To provide shared language, the College proposes three AI modes for assignments.
- AI-0 (AI Prohibited): No AI can be used to complete this assignment, including for search, code and text generation, debugging, and calculation.
- AI-1 (AI Permitted): AI tools can be used on this assignment in a way that supports the learning objectives and does not short-circuit students’ productive struggle. Faculty are encouraged to give specific examples of what effective and ethical AI usage looks like on a specific assignment or set of similar assignments.
- AI-2 (AI Required): AI tools must be used to complete this assignment. These assignments will often be ones designed to help students learn to use AI.
- Faculty should choose the AI mode for each assignment in order to support the course’s learning objectives. We generally expect courses to include a mix of AI-0, AI-1, and AI-2 assignments. AI-1 and AI-2 assignments can often take on more advanced, complex work than assignments did in the pre-AI era. But faculty should avoid unrealistic trick problems and instead focus on complex, applied challenges drawn from the real word. These are the problems that help students develop the intuition to check AI outputs appropriately.
- Syllabi should include a section titled “AI-Independent Concepts and Skills” that tells students what they should be able to explain or do without AI tools by the end of the course. Examples might include explaining the second law of thermodynamics and why it matters or setting up the equations of static equilibrium for a given problem. These expectations should be clearly communicated in the syllabus and reinforced in class so that students recognize the need to practice these concepts and skills independently well before their first AI-0 exam.
- Faculty should allow students to use AI to prepare for AI-0 exams, like how they are allowed to work with other students to prepare for closed-book exams.
- Design over policing: Assignments should be designed to support the assigned AI mode without substantial policing (e.g., an AI-0 assignment is best delivered in-person and on paper or using the Lockdown Browser with video). For AI-1 and AI-2 assignments, faculty are encouraged to ask students to disclose how they used AI, both to promote student accountability for beneficial AI use and to give instructors data.
- AI-use disclosure: For AI-1 and AI-2 assignments, faculty are encouraged to ask students to disclose how they used AI. This can promote accountability for beneficial AI use and help instructors identify concepts students are struggling with. Faculty may require one of three forms of disclosure:
- A transcript of the AI interaction
- A student-written summary describing how AI was used and what work the student contributed
- A summary generated by the AI tool in response to a prompt provided by the instructor
- Ethical use and data integrity: Early in each course, instructors should include an assignment that helps students understand what constitutes ethical use of AI, which data sources they can legally and appropriately use, and common missteps to avoid. The assignment should be tailored to the course content.
- Homework: Making homework AI-0 is not recommended. Instead, you may consider giving AI-1 homework that is worth a small portion of the course grade and following each assignment with an AI-0 in-class quiz, written or oral, based on one of the homework questions.
- In-class practice: Consider incorporating at least one AI-0 warm-up problem in each class meeting. This could be a five-minute problem, where the student just sets up the approach – i.e., they don’t have to fully complete the subsequent computation. Consider having a student who did well present 4 their answer in the following class (with your commentary). This ensures a baseline level of productive struggle for all students and gives them exposure to a range of problem-solving approaches.
- Encourage students to share their AI usage ideas with you: Faculty are encouraged to ask students to share when they find an AI tool or workflow that helps them with the class. This allows faculty to learn and check with topical expertise and helps create a collaborative relationship between faculty and students related to AI.
- Provide a custom Gemini Gem for your course: Create a custom Gemini Gem for your course that has all your material plus custom instructions. Provide the Gem to your students at the start of the semester along with guidance for beneficial use. This is more efficient than all students creating their own Gems and allows you to provide a Gem that is tailored to your specific approach and preferences. In your Gem instructions, include instructions for the Gem to issue a warning when a student asks a question about something that Gemini typically gets wrong. We are recommending Gemini currently because all students and faculty have accounts through the university. You can learn how to create a Gemini Gem here. Be sure to use your UH Gemini account for this.
- Use AI to proofread assignment instructions: Faculty should run draft assignments through AI with a prompt like, “Please flag and suggest edits where these instructions may be unclear to a student at a 10th-grade reading level who may be a non-native English speaker.” This is a use of AI that can reduce a lot of unhelpful friction students face and is well-suited to generative AI’s strengths.
Learning to use AI
It is an instructor’s responsibility to have an up-to-date understanding of how AI is used in practice in their course content area. When it is clear that a particular AI tool or workflow is commonplace in engineering practice related to the subject you are teaching, you should teach students how to use that tool or workflow. A useful example: engineering design courses had to change when CAD software became standard practice. To properly prepare students, courses had to incorporate explicit instruction in and practice with CAD. You are encouraged to work with your department and draw on university resources like TIP grants to fund your own learning and experimentation with evolving tools related to the subject you are teaching. You are also encouraged to stay in touch with a broad range of professional and academic engineers/technologists to develop a good sense of how AI is being used in practice. Attending our biannual career fair before it begins to talk to on-site engineers/technologists who recruit our students is an easy way to do this regularly. It takes just one hour each semester.
Additionally, faculty are expected to learn and understand key limitations of AI related to their course content and communicate these limitations to their students. For example: in 2025, the free version of ChatGPT could not produce an accurate free-body diagram. Statics professors should know this and communicate it to their students to prevent the non-beneficial use of AI. Staying up to date requires that instructors test the best model that our students have access to on their course material every semester.
Finally, we recommend that all departmental and unit chairs appoint an AI ambassador who will catalyze discussion within the department about best practices, assist with technical mastery, and, in general, serve as a resource.
Guidance for Students
Using AI to learn
As a student, your primary goal is to learn and build skills that you will use for the rest of your life. Learning requires both high-quality learning materials and productive struggle. What makes AI challenging is that it can amplify both of those necessary ingredients for learning, but it can also remove them entirely.
Generally, you will see that AI-mode is assigned at the assignment level. It is your responsibility to know and follow the AI mode for each assignment.
- AI-0 (AI Prohibited): No AI can be used to complete this assignment, including for search, code and text generation, debugging, and calculation.
- AI-1 (AI Permitted): AI tools can be used on this assignment in a way that supports the learning objectives and does not short-circuit students’ productive struggle. Faculty are encouraged to give specific examples of what effective and ethical AI usage looks like on a specific assignment or set of similar assignments.
- AI-2 (AI Required): AI tools must be used to complete this assignment. These assignments will often be ones designed to help students learn to use AI.
For AI-1 assignments, a good check is - could I explain the underlying concepts and my complete approach to a friend or family member who has no context on this subject? Including the key assumptions and decisions I made in setting up the problem and why they matter? If the answer is not a confident yes, you have more work to do, even if your assignment looks complete.
You are not expected to develop this practice alone. Your instructors will provide guidance and you are encouraged to ask questions and engage in dialogue with them when you aren’t sure what beneficial, ethical AI usage looks like for a particular assignment. At the end of these guidelines, you will find example AI workflows that can facilitate your learning goals.
Learning to use AI
As an engineer or technologist, you are expected to have the “ability to acquire and apply new knowledge as needed, using appropriate learning strategies. (ABET Student Outcome 7).” The rapid development of AI technology while you are in school offers a great opportunity to cultivate this critical ability. Some courses will have explicit AI-based modules while others will not. Through these modules and your own experimentation with AI through the completion of AI-1 and AI-2 assignments, you should learn these new tools and observe how you update your practice as they evolve.
Students are encouraged to develop a step-by-step approach to their use of AI. A useful framework called TACO was developed by Cecilia Ka Yuk Chan (2026) and is outlined here:
| Phase | Core Purpose | What the Learner Does | What It Prevents |
|---|---|---|---|
| T—THINK | Human-first cognitive initiation |
|
Premature outsourcing; starting with AI instead of thinking |
| A—ASK | Bounded and intentional delegation |
|
Unrestricted delegation; AI doing the entire task |
| С— СНHECK | Verification and epistemic control |
|
Blind acceptance; fluency bias; misinformation |
| O—OWN | Internalisation and accountability |
|
Performance without understanding; externalised cognition |
An important part of using AI responsibly is understanding privacy and data integrity risks. We recommend using AI tools through university-provided accounts (Google Gemini and Microsoft Copilot). If you use other AI tools, make sure you have disabled any settings that allow your inputs to be used to train the company's models. Never upload private or confidential information, or data you do not have permission to share to these unauthorized tools.
Learning to use AI also means understanding how it works. Engineers and technologists are curious about how things work, and by graduation, you should have a basic understanding of how generative AI models are built, trained, and tested.
Research, Dissertations/Theses, and Graduate Students
The principles and coursework guidance above also apply to graduate education. The following additional guidance applies to all students engaged in research, graduate students primarily but also including undergraduate researchers and students preparing theses or dissertations.
- Responsible use and independent scholarship: Students are encouraged to learn how AI can support research, in consultation with their research advisor. Advisors should discuss appropriate uses and expectations early in the project. Students must develop the knowledge and judgment needed to explain and defend their methods, results, and conclusions, and demonstrate the intellectual contribution required for their degree.
- Human accountability: Students are primarily responsible for the accuracy, originality, and integrity of their work, including AI-assisted writing, calculations, code, analyses, figures, and references. AI outputs require critical review and verification. AI tools cannot be listed as authors, consistent with publisher policies such as Elsevier’s. Fabricated citations, invented evidence, and misleading 7 representations of research are unacceptable regardless of the tools used. Faculty must ensure that these guidelines are carried out.
- Disclosure and documentation: We encourage open discussion and transparent disclosure of AI use as part of responsible scholarship. Theses, dissertations, and other research outputs should disclose substantive AI assistance, including assistance with scientific reasoning, analysis, coding, interpretation, or drafting. Disclosures should identify the tool and version, explain how it was used, and describe how its outputs were checked. AI used as part of the research methodology should be documented in the methods section with sufficient detail to support reproducibility. Students should retain relevant records and, when uncertain whether disclosure is needed, favor disclosure.
- Journal and publisher requirements: Students and advisors must follow the current AI policies of their intended journal or publisher. For example, the American Physical Society (APS) requires disclosure of substantive AI use, while exempting routine polishing or light editing. Requirements differ across publishers and may change; students should check them when planning research and again before submission. Publication requirements supplement applicable University and graduateprogram requirements.
- Protection of research information: Use of AI must comply with the University policies and applicable research agreements concerning confidentiality, privacy, intellectual property, and data security. Confidential or restricted research material should be entered only into tools approved for that purpose.
Frequently Asked Questions
Is creating a course-specific Gemini Gem required, and how does this recommendation align with UH’s privacy and IT policies?
Creating a Gemini Gem for your course is a recommendation for Cullen College faculty, rather than a university- or college-wide requirement. Some students are creating their own Gems for each course. Providing a course Gem and sharing it with students allows instructors to select appropriate source materials and customize its guidance—for example, by highlighting common misconceptions and areas where AI tools tend to make mistakes. We provide the following language that could be included on your syllabus or Canvas with a link to the Gem.
A custom Gemini Gem for this course is available to help you review concepts, ask questions, and practice your skills. Access it using your UH Google Workspace account. The Gem is a supplement to, not a replacement for, course materials, in-class instruction, and office hours. It can make mistakes—even when its answers sound confident—so check its responses against course materials and ask your instructor when something is unclear or seems off.
UH provides privacy-protected access to Gemini for faculty and staff through university Google Workspace accounts. The Gemini licenses at UH have been approved for Level 1 data protection by both UIT Security and UH Legal. So, as long as the UH licenses are being used, there are no data issues. Data uploaded through this account is not shared or used to train AI models. Please see ai@UH’s AI Tools page for additional information and training.
Are there guidelines for using AI to grade?
The overarching guideline for grading is that the instructor is accountable for accurate, timely, and helpful grading. Whether to use AI for grading can be determined at the department or individual instructor level.
Drafting and AI-Use Disclosure
The initial draft was written by a faculty member without using AI. It drew on discussions with students and colleagues, the August 2026 Cullen College Roundtable on AI and Critical Thinking, and the January 2026 Faculty AI Workshop. After the draft was completed, ChatGPT (GPT-6.0 Astra) was used to compare it with AI policies from other engineering colleges, assess its alignment with the College’s AI philosophy, and identify typos. The document was then revised in response to feedback from the Cullen College Executive Council and faculty across the college.
Download “Guidelines for the Use of AI at the Cullen College of Engineering” (PDF)