Student-facing guidelines and policy: Guidelines on using Generative Artificial Intelligence (GenAI) for Coursework (Undergraduates)

Staff-facing guidelines and policy: Using Generative AI at Dartmouth - Staff Overview

These assignments were redesigned under the GenAI Teaching Grant and are shared with the grantees' permission:

• Chad Elias, Art History
  Comparing Human and AI Interpretations of Visual Art. Art history students were asked to first conduct a close visual analysis of a photograph by artist Carrie Mae Weems from her series Repeating the Obvious (2019)—a blurred image of a young man wearing a hoodie—and then prompt an AI chatbot to analyze the same image. By comparing human and AI generated interpretations, students examined what these systems foreground versus what human viewers perceive, particularly when analyzing artworks that engage questions of representation, visibility, and implicit bias. Their essays documented both analyses and reflected on the limits of machine interpretation, asking whether machine-learning systems can meaningfully account for cultural and historical context in socially engaged art, and how AI may reproduce, obscure, or reframe latent and deep-seated assumptions brought to the act of visual interpretation.
• Christopher Sneddon, Geography and Environmental Studies
  Using GenAI to Research NGOs, Local Viewpoints, and Compare Information Sources. Geography/environmental studies students used GenAI across three linked tasks exploring environmental politics in Southeast Asia. First, they prompted AI to generate lists of regional NGOs (specifically excluding Western organizations with in-region programming). Second, they created summary reports on selected NGOs—covering activities, tactics, constraints, and political-economic contexts—while documenting how they integrated GenAI into their research process. Third, they prepared for a role-playing exercise on Lower Mekong Basin dam construction using both traditional search engines and GenAI tools, then compared the relative strengths and weaknesses of each approach in providing compelling arguments and accurate information. This scaffolded assignment built both regional expertise and AI literacy.
• Elisabeth Newton, Physics and Astronomy
  GenAI-Assisted Science Art Projects with Clear Boundaries. For their final project, Astronomy 1 students could either participate in a citizen science project or create an artistic project that makes astronomical concepts accessible to a general audience. Students could choose to use GenAI as a creative tool, but not as a substitute for their own work. Clear examples distinguished appropriate from inappropriate GenAI use: For example, prompting AI to write a complete song about Pluto was unacceptable, while using AI to generate illustrations for a student-written children's book (with the student responsible for ensuring the images' accuracy) was allowed. Students who chose the creative project submitted documentation in stages—brainstorming notes, proposal/outline, and scientific explanations—giving Professor Newton insight into their processes and (if used) AI integration.
• Using AI to Analyze Academic Articles and Model Energy-Related Policy Perspectives. Energy and Environment students compared AI-generated article summaries with their own analyses, experimenting with prompting strategies to evaluate output quality and accuracy. An A/B analysis revealed that students who read articles before using GenAI developed a much deeper understanding of AI's limitations than those who used AI first. Students also used GenAI to simulate positions of various stakeholders concerning specific energy-related executive orders. This involved analyzing published statements and generating content such as tweets, press releases, and policy briefs to understand stakeholder sentiments and priorities. Finally, they used AI tools to analyze peer-reviewed literature and stakeholder perspectives while developing their own policy briefs.
• Eugene Korsunskiy, Engineering
  Demystifying GenAI Through Bounded Experimentation in Design Thinking. Students in a Design Thinking engineering course were asked to imagine the ways that AI-enabled interactive digital technology might enhance the human experience by the year 2032—and to use GenAI in their design process. Following a prompt-engineering workshop, students received specific guidelines delineating appropriate and inappropriate GenAI use at each project stage. Many students initially viewed any GenAI use as "cheating," but the combination of carefully established guardrails and experimentation helped demystify the technology. Students reported hope in GenAI's capacity as a tool to assist them in developing novel ideas and solutions, particularly valuing its ability to jumpstart brainstorming sessions when they felt stuck. This structured approach helped students distinguish between using AI as a creative catalyst versus an academic shortcut, giving them a greater understanding of AI's role in design thinking. If any Dartmouth faculty member would like to see the complete assignment instructions, feel free to email Eugene.
• Brian O'Connor, Institute for Writing and Rhetoric
  AI That Only Asks Questions to Support Close Reading. Writing 2/3 students used GenAI as a Socratic questioner rather than answer-generator. The professor created a structured prompt that instructed AI to push students toward deeper textual understanding through questions alone, never providing analysis. This judgment-free practice space supported close reading skills—rereading, reflection, interpretation—essential to first-year writing, transforming AI from shortcut to learning scaffold. This was the only permitted GenAI use in the course.
• Nikhil Singh, Computer Science
  Future GenAI Developers Analyze and Reflect on Their Own AI Interactions. Computer science students in a Human-Centered Generative AI course analyzed their own interactions with AI systems through minimally structured reflective assignments. Rather than prescribing specific procedures, Singh asked students to engage with at least one GenAI system—from ChatGPT to GitHub Copilot to multimodal models—and connect their hands-on experience to course readings on ethics, design, and human-AI collaboration. Student reflections surfaced critical themes including creativity versus homogenization, active challenging versus passive acceptance of outputs, cognitive load from unlimited generation, and uncertainties around copyright and analytical reliability. The assignment prompted deeper discussions about what constitutes skillful writing when fluency is easily achieved using automation, leading them to recognize that "the new signals might be deviation from the patterns LLMs implement by default." By moving students from spectator to participant, the course developed critical GenAI literacy essential for future AI developers.
• Exploring Technical and Policy Documents for Alternative Energy Projects. Offshore Wind Power students used GenAI to summarize lengthy documents (often exceeding 2,000 pages)—including government reports, environmental impact statements, records of decision, procurement orders, and public comments—for integration into project Wiki pages. They also used GenAI as part of an in-class activity analyzing how different stakeholder groups might react to executive orders targeting offshore wind development. Students regularly queried industry visitors/speakers about workplace GenAI use and anticipated developments. 
• Jonathan Chipman, Geography
  From Manual Map Coding to AI-Assisted Geovisualization. Geovisualization students used GenAI tools to write JavaScript code for interactive web-based maps. After initially coding by hand in Leaflet.js to understand fundamental mechanics, students progressed to using GenAI for two purposes: solving specific, narrow coding problems and generating complete web map applications from scratch. This required learning to formulate effective prompts, identify issues, iterate solutions, and critically evaluate AI outputs. The course integrated discussion of social context and ethical issues surrounding AI use, balancing technical proficiency with critical literacy in automated coding. If any Dartmouth faculty members are thinking about adopting AI for use with coding in their classes, feel free to email Jonathan.
• Using GenAI for Analyzing Non-English Scholarly Texts. Interdisciplinary studies students piloted GenAI tools for content analysis of a Spanish-language journal dataset, supporting faculty research on how academic fields evolve outside English-speaking contexts. After hands-on training with a data science specialist, students used GenAI tools to generate summaries of individual articles and answer course-related questions. Graded on completion to encourage honest experimentation, this low-stakes assignment revealed that while LLMs effectively summarized texts, they were poor substitutes for close reading, with specific weaknesses in conceptual associations and knowledge of liberation theology, indigeneity, and LGBTQ+ topics. Technical constraints (token limits, inferior free models) became lessons about AI infrastructure's impact on research. One Spanish-language learner benefited significantly from AI as a language-translation tool. The professor plans to scaffold prompting skills earlier, teach students to fine-tune model settings, focus retrieval on specific document sections, and systematize student feedback for future iterations.

Note: The introductory titles for these summaries were developed with assistance from Claude Sonnet 4.5, accessed through Dartmouth Chat.

The following examples were drawn from Dartmouth and other institutions. Examples from outside Dartmouth are sourced from Lance Eaton's Syllabi Policies on Generative AI, a crowdsourced repository that is regularly updated with new contributions.

The statements below are organized alphabetically by discipline and identified by faculty name and institution. To have your syllabus statement added to this collection, please share it with us via this form and indicate whether you would prefer to share anonymously.

Anthropology Course

Charis Boke, Dartmouth College

You are expected to develop and maintain a thoughtful relationship with tools of artificial intelligence which can support writing and creating. We will be working directly with AI writing tools several times through the semester, exploring their opportunities and weaknesses for Environmental Justice studies, and citing them appropriately. However, I ask that you do not use AI generators to create work unless specifically requested to. We will learn together. 

Art History Course

Mary Coffey, Dartmouth College

Academic Honesty and Integrity:

It is impossible to create a supportive and collaborative learning environment when members of the community are being dishonest about their work. Dishonesty can take many forms from cheating, over-relying on AI writing supports, exploiting classmates work ethic, coming to class consistently under-prepared (coasting), skipping class, or failing to communicate when something is impeding your ability to live up to your obligations to the class (ghosting). I take honesty and integrity VERY seriously, especially now that we are living in a culture that celebrates and rewards dishonesty. Integrity requires consistent effort, self-reflection, and accountability. I hold myself to these high standards, and I expect the same from my students. The rules are always changing as new technologies come online, as pre-college education is more and more impacted by defunding and politicized attacks, and as younger generations bring new skills, challenges, and needs to higher education. My policies are based on Dartmouth College's honor code, over 30 years of undergraduate teaching, and my experiences as a parent to gen-z and gen alpha learners. They evolve and change, and your honest input can impact how they do.

AI Writing Supports:

I consider all AI writing supports (with the exception of simple spell and grammar checks available for free in all word-processing programs) as short-circuiting the necessary relationship between thinking, writing, and learning. While you may have professors that encourage you to use AI or you may believe that you can use AI effectively, my experience in the classroom has shown me that very few students use it as a learning tool and too many rely on it to evade the hard work of putting their thoughts into language, organizing their arguments, or refining their writing skills and writerly voice.

• AI produces banal texts that are often repetitive and riddled with inaccuracies.
• AI writing tools plagiarize the work of scholars without their consent and without any concern for citation.
• AI "smooths" writing, eliminating the unique voice of individuals and replacing it with an institutionalized voice that reinforces the idea that "good" writing is affectively "neutral."
• AI does not permit authors to learn from their writing errors or to engage in the kind of sentence-level writing and revision that is essential for developing skill as a writer and engaging readers.

In every way, AI writing supports contravene the learning objectives and pedagogical design of this course. For that reason, I do not permit AI writing supports in this class.

I make a distinction between AI writing supports and AI reading and research supports. It is permissible to use, for example, google translate—in this class only—to read articles originally published in Spanish. You will be relying on algorithms in online or library search engines, for example. But you may not ask ChatGPT or any other AI program to generate a bibliography for you (this is perilous as AI will invent sources). You may not use grammerly's AI function or ChatGPT or any other LLM to edit your writing, fix your grammar and spelling errors, etc. You may use word or google doc's basic spell and grammar checks. Those tools merely identify misspelled words or grammatical errors and offer you the option to fix them. In some cases, the checks are wrong, which is why it is important for you to use your own discretion when accepting or rejecting editorial suggestions generated by AI. If you are using an AI program that I have not mentioned here, you MUST not use it in this class unless you inquire with me beforehand about whether it falls under the permissible or not permissible uses of AI in this class. All classes are different. What I may accept, another professor might not and vice versa. It is your responsibility to do due diligence on this matter. Ignorance is not an excuse.

If you want to try to persuade me to modify my policy or to discuss a particular use of AI, I welcome an in-person meeting during office hours or at a time convenient for us both. However, this kind of conversation must take place BEFORE you have violated the policy. When in doubt, ASK first. Permission will not be granted ex post facto.

If I determine that you have used AI writing supports of any kind, I will first call you in for a conference to discuss my concerns and to determine the consequences which can range from failure of the individual assignment to failure of the class and a formal report to the Committee on Standards which can result in separation from the college. These are not idle threats. I failed two students in Winter 2024 for violating my policies. For reference, in my 20 years teaching at the college I had only failed 2 students before last winter.

Cheating:

It goes without saying that submitting work, of ANY KIND, that is not your own is CHEATING and will result in failing the class.

Students who make honest mistakes with citation or who confuse plagiarism for paraphrasing will be given the opportunity to learn from their mistakes the first time. After that, violations will be considered intentional and will result in failing the class.

Cultural Studies Course

Sarah Bunin Benor, University of Southern California

ChatGPT and other AI generators that use large language models can be useful for researching and writing papers. However, you should be aware of their limitations:

1. Errors: AI generators make mistakes. Assume the output is incorrect unless you check the claims with reliable sources.
2. Bias: Their output may reflect bias because the data they are trained on may reflect bias or may not include sufficient data from certain groups.
3. Citation: These tools use existing sources without citation. Therefore using their outputs puts you at risk of plagiarism.

With these limitations in mind, you are welcome to use AI generators to brainstorm and refine ideas, find reliable sources, outline, check grammar, refine wording, and format bibliographies. Beyond bibliographies, you are not allowed to copy and paste material generated by AI and use it in your assignments. At the end of your bibliography, add a note indicating which AI tool you used and how you used it, including the prompt(s) you used and the date(s).

Engineering Course

Kate Goodman, University of Colorado Denver

Utilizing ChatGPT or other AI tools is becoming more common. While I would prefer you not use these tools and instead commit to the productive struggle that is learning, I recognize that these tools are not going away. Rather than ban them, we will treat them similarly to other resources you use. This means you MUST follow these four points: 

1. Give notice that you used the AI tool, which one you used and how you used it in the comments of your code. 
2. Rigorously test and alter the program to suit the assignment and your understanding. 
3. You must understand any code you submit and be prepared to explain it to me.
4. All comments should be your own words.  Sample code with the appropriate credit statement will be shown in class.

English Course

Nirvana Tanoukhi et al., Dartmouth College

Use of Generative Artificial Intelligence (GenAI). We are still in the early stages of learning to navigate GenAI technologies, and new tools will continue to become available. With this in mind, the following course policies are provisional and subject to change. 

First, some words of caution. While GenAI has shown remarkable potential as a supplementary tool for brainstorming, thinking, writing, and revising, there are many things it cannot do. There are also real downsides to over-relying on it.

Notably, GenAI works by text prediction. This means that by design, it tends toward unoriginality and even cliché. It also means that GenAI is not beholden to the truth. When a chatbot delivers a response to your prompt, it is telling you something that might sound right, not something that it has vetted for accuracy.

Furthermore, even if a chatbot, working on its own, could produce a perfect, A+ essay for you (it can't), something would be lost in this transaction. At its best, English homework is designed to develop your skills of careful observation, creative and experimental thinking, nuanced analysis, and authentic self-expression. It is designed as an occasion for learning. If you outsource your homework to a chatbot, you will risk diminishing your own learning experience.

Here, then, are our rules:

1. Use of GenAI on written assignments is permitted at your discretion, provided that it is judiciously implemented and reviewed, and properly documented at the time of submission. For any assignment on which you use GenAI, you must turn in a cover letter, including an explanation of your strategy and reasoning for using the technology (one to two paragraphs will suffice), a comprehensive and verbatim list of the prompts you used, and a note on how you checked the accuracy of the output (another paragraph here).
2. You are responsible for what you turn in for assessment, including any inaccuracies or factual errors in the text. 

If you are uncertain about whether a particular application of GenAI complies with our course policies, or if you have questions or concerns that are not anticipated here, please get in touch with me. I welcome your thoughts and will have much to learn from your experiences with this technology as it evolves.

First-Year Writing Course

Loretta Notareschi, Regis University

As scholars, we have an obligation to share with our readers the sources and tools we used in creating our scholarship. This is both because it is dishonest to portray other people's ideas as our own and because it is helpful to our audience to put our work in the context of the greater scholarly conversation. Readers may be curious to learn more about our subject; they may want to verify our information; or they may even want to create their own scholarship inspired by ours. In all cases, they will need to know what our sources were. To this end, every paper should have two features indicating our reliance on outside sources:

The first should be in-text parenthetical citation paired with a Works Cited list (in APA or MLA style); or Footnotes/Endnotes and Bibliography (in Chicago Style) with the authors, titles, publishers, dates, and URLs (if appropriate) of each source. This is for sources we have quoted directly (which should be in quotation marks), those we have paraphrased in our own words, and those that we have used for background information. All sources for the text should be properly introduced, with their connection to our own ideas clearly stated. 

The second should be an Artificial Intelligence Disclosure, which should contain the following statements:

"I did not use artificial intelligence in creating this paper" or "I did use artificial intelligence in creating this paper, namely ____________ (ChatGPT, Bard, etc.). I used it in the following ways (check which of the following acceptable uses were utilized):

• Brainstorming help
• Outlining help
• Background information
• Grammar/spelling/punctuation/mechanics help

and I affirm I did not generate text with artificial intelligence and directly copy it into my paper."

Why is it important not to directly copy words from an AI engine into our texts? There are multiple reasons: first, this would be considered plagiarism (which means presenting others' words as if they were our own); second, AI engines are notoriously unreliable on facts—anything they assert must be checked against reliable sources; third, AI engines reproduce biases and prejudices from their source material—it is incumbent on us to check and correct for bias; and finally, using AI to generate text may rob us of the chance to develop our own thinking on a subject. Think about it this way: the point in education is not to generate text artifacts. Rather, the point is to help us develop our own ability to think critically. Writing is a means to critical thinking, and we must do our own writing to cultivate our own true, not artificial, intelligence.

Literature Course

Alexa Alice Joubin, George Washington University

Using an AI-content generator such as ChatGPT to complete assignments without proper attribution violates academic integrity. By submitting assignments in this class, you pledge to affirm that they are your own work and you attribute use of any tools and sources. 

Learning to use AI responsibly and ethically is an important skill in today's society. Be aware of the limits of conversational, generative AI tools such as ChatGPT. 

• Quality of your prompts: The quality of its output directly correlates to the quality of your input. Master "prompt engineering" by refining your prompts in order to get good outcomes. 
• Fact-check all of the AI outputs. Assume it is wrong unless you cross-check the claims with reliable sources. The current AI models will confidently reassert factual errors. You will be responsible for any errors or omissions.
• Full disclosure: Like any other tool, the use of AI should be acknowledged. At the end of your assignment, write a short paragraph to explain which AI tool and how you used it, if applicable. Include the prompts you used to get the results. Failure to do so is in violation of academic integrity policies. If you merely use the instructional AI embedded within Packback, no disclosure is needed. That is a pre-authorized tool.

Here are approved uses of AI in this course. You can take advantage of a generative AI to: 

• Fine tune your research questions by using this tool https://labs.packback.co/question/  Enter a draft research question. The tool can help you find related, open-ended questions
• Brainstorm and fine tune your ideas; use AI to draft an outline to clarify your thoughts
• Check grammar, rigor, and style; help you find an expression

Life Sciences Course

Franklin Hays, University of Oklahoma

Use of AI tools (e.g., ChatGPT, Bard, Claude) are encouraged in this course to facilitate the student learning experience and overall productivity. However, such use should follow three clear principles: 1) any and all use should be transparent, properly cited, and otherwise declared in any final work product produced for grading or credit; 2) students are responsible for ensuring accuracy of content produced including references and citations; and 3) students acknowledge that improper attribution or authorization is a form of academic dishonesty and subject to the Academic Misconduct Code as outlined in the Student Handbook and the Faculty Handbook. All work turned into the instructor for grading is assumed to be original unless otherwise identified and cited. If there is uncertainty about any content in regard to the above guidelines, please contact the instructor to discuss these questions prior to turning anything in for grading.