Harvard releases a guide to generative AI student applications with four recommendations and 33 strategies

On January 17, 2025, the Harvard Graduate School of Education (HGSE) released the guide "GenAI in Student-Directed Projects: Advice and Insights," which was developed by the Harvard Creative Computing Lab based on the experiences of Learning Design ( The guide, written by the Harvard Creative Computing Lab based on the experiences of Learning Design students, demonstrates effective ways that GenAI can complement teaching and learning in the context of supporting creative student-directed projects that focus on student autonomy, critical thinking, and problem solving. The guide emphasizes the opportunities and difficulties presented by the new technology and provides recommendations, strategies, and some things to keep in mind when using it.

Karen Brennan, director of the Creative Computing Lab, co-author of the guide, explains, "In K-12 and higher education, part of the anxiety about generative AI and its potential misuse stems from the realization that we may be asking learners to do work that isn't actually particularly important or meaningful. If machines can do this, what does it mean to ask students of all ages to do it?"

Original text:https://creativecomputing.gse.harvard.edu/genai/

Karen Brennan, Paulina Haduong, Avantika Kolluru, Sally Yao, Jacob Wolf
Harvard University Graduate School of Education

introductory

"The question is not whether the AI is more powerful than you, but whether you plus the AI are more powerful than you."

- Andrew Ho, Professor, Harvard University Graduate School of Education

Student-led projects, i.e., projects in which students are guided in content and process, can serve as personalized and meaningful contexts for the development of a variety of skills, concepts, and disciplinary experiences and fluencies. However, students may experience difficulty in defining goals, determining the scope and tasks of the project, and concretely realizing their creative vision.

"Student-led" does not mean completely autonomous work. In fact, students can benefit from a variety of scaffolding and support when working on autonomous projects. At this juncture, Generative Artificial Intelligence (GenAI) is becoming more accessible as a new form of support. When we talk about GenAI, we mean tools like OpenAI's ChatGPT or Harvard's customized AI Sandbox, which generate text, images, or other content on cue.

Based on our own experiences as teachers and students, we were curious about how GenAI could serve as an effective support in student-led programs. This guide, generously supported by the Harvard Initiative for Learning and Teaching (HILT), is the result of our exploration of that curiosity.

This guide contains advice and insights on how GenAI can support self-directed projects that support students to engage in non-prescriptive work in the face of uncertainty and opportunity, where the outcome is neither guaranteed nor known. It is our hope that this guide will support students and teachers in exploring how GenAI can be used to realize students' goals and aspirations.

Our Process

In Fall 2023, we taught a core required course for Instructional Design students at the Harvard Graduate School of Education. Students in this course developed a semester-long independent project focused on designing a learning experience. Each student participated in a project - but the projects they worked on and how they conducted them were up to them.

With the new focus on GenAI, we wanted to understand how students in the program are using these tools in program development. We were also curious about the reflections of colleagues teaching program courses on the challenges and opportunities of this particular technological moment. We interviewed 27 Harvard Graduate School of Education instructional design students and seven Harvard Graduate School of Education faculty members.

Prior to the student interviews, we asked each student to prepare a map of the project development experience, marking key milestones along the journey and pointing out important supports and tools along the way, including GenAI.This process orientation helped us to understand the various activities involved in the multi-month Learning Design project.

In addition to the process questions, we asked students to share their broader thoughts about the autonomous program and GenAI. The faculty interviews were similarly structured, but focused on their students' work rather than their own use.

Of course, these interviews reflect a particular point in time and a particular group of students. With regard to a specific point in time, we wanted this record to serve as a record of the possibilities people feel when they begin to engage with powerful new technologies. Regarding a specific group of students, we worked exclusively with graduate students studying design. But even if you don't work with graduate students or don't focus on design for learning in your teaching, we hope that ideas about creative, student-led design projects will still inspire your teaching and design imagination.

Recommendations and Implications

In our conversations with students and faculty, we have heard many wonderful general suggestions and beautiful specific examples of using GenAI in student-led programs. We've written this guide to share the breadth and depth of what we've learned. First, the general advice:

Consider the broader implications. Students and faculty emphasized the importance of thoughtful use, pointing out the many serious issues associated with the use of GenAI, ranging from its tendency to produce hallucinations, its large environmental footprint, accessibility barriers due to cost, potential homogenization of cultures, and inherent algorithmic bias. The advice given is clear: approach GenAI with intent and awareness of its limitations and potential hazards.

Preserving learning and authentic voices. Another recurring theme was the use of GenAI as a support rather than a substitute for individual thought, effort and style. As one student explained, "I think it can be your second brain if you're feeling a bit stuck. You can ask it to help you get started, but it can't do everything for you. You're still the pilot, it's just an assistant." Students described the challenges of deciding which tasks to delegate to GenAI and which to tackle on their own. As another student suggested, "Really think about what you want at this moment. Do you just want to get the job done, or do you want to learn?"

Embrace play and experimentation. Students and faculty alike emphasized the importance of hands-on exploration across different GenAI tools. "Make sure you use it and explore it," one student advised, while noting, "It can't do everything for you. So use it strategically." Faculty echoed this sentiment, with one faculty member saying, "You've got to play with it and get a feel for it. You can't help students until you feel it yourself."

Practice strategic iteration. Students shared that they have learned through experience that success with GenAI requires multiple attempts and prompts for improvement. As one student observed, it "doesn't give you the right answer immediately after the first try, you have to modify it a little bit until you get the answer." However, they also warned of the learning curve associated with new tools, recognizing that sometimes existing methods may be more effective. As one student wisely suggests, after recounting a particularly lengthy GenAI experience, "If you realize you've spent six hours on it, maybe you should be done by now, maybe you should stop."

These guiding suggestions are accompanied by specific examples to inspire practical action. The remainder of this guide provides a collection of strategies for using GenAI in student-led programs.

The order in which these strategies were presented was inspired by the sequence of activities we saw and heard in the student maps. We heard about the early experiences of their projects, developing ideas about the project space and exploring project possibilities. We heard about the long and messy intermediate stages of creating prototypes as tangible manifestations of thinking, sharing them with others, and as part of an iterative process of revision and refinement. These strategies were used throughout the process.

Each strategy is accompanied by a short title, a short description, a student reflection, and examples of the strategy in action.

Examples of Strategies

Examples of strategies in action include project overviews and tips on GenAI tools that may support project work. All projects and tips are inspired by actual student projects. The prompts in this guide are used to generate output from the GenAI tool, but the tool is in continuous development; if you try a prompt in the same tool today, the output may vary.

Where possible, we have only included examples of tools that offer a free tier, but there are various pricing models for GenAI tools. We hope this guide inspires you on how to use GenAI as an additional resource for your own autonomous projects or to support the autonomous projects of others.

1. Starting point: information gathering and research

Synopsis. Use Generative Artificial Intelligence (GenAI) as a starting point for gathering information. For example, students could ask GenAI for information about the upcoming solar eclipse and then look further into other resources, such as YouTube videos or academic papers, as needed.

2. Brainstorming: developing project ideas

Synopsis. Use GenAI for conversational brainstorming to spark project ideas. For example, students can discuss with ChatGPT how to build an online platform for sharing campus stories, getting advice on platform features and technical implementation.

3. Building Chaos: Developing a Project Plan

Synopsis. Use GenAI to generate a project timeline with goals, activities, and deliverables. For example, students can ask GenAI how to break down a six-week museum exhibit project into manageable steps and tasks.

4. Define it: explore related concepts

Synopsis. Use GenAI to gain a deeper understanding of project-related concepts or constructs. For example, students can ask GenAI about the definition of "student autonomy," its characteristics, and the relationship between those characteristics to better understand the research topic.

5. Persona possibilities: analyzing target audience needs

Synopsis. Use GenAI to generate user personas to explore the latent needs of your target audience. For example, students can ask GenAI to create user personas for a music learning app for adult learners, including information on demographics, goals and motivations, pain points, learning preferences, and more.

6. Context matters: applying concepts to relevant contexts

Synopsis. Apply concepts and ideas to personally relevant contexts. For example, students could ask GenAI about policy differences in promoting makerspaces in India and request relevant data sources.

7. Scope and size: defining project constraints

Synopsis. Focus on key elements by defining project constraints. For example, students could ask GenAI about what constraints should be set for organizing an event on climate adaptation careers, including the length of the event, the deliverables, and what is out of scope.

8. Bold blueprint: developing a strategic plan

Synopsis. Create a strategic plan based on vision and goals. For example, students could ask GenAI to provide a draft strategic plan, including a SWOT analysis and strategic goals, based on a nonprofit's mission statement and general job description.

9. Conceptual analogies: explaining complex concepts

Synopsis. Explain difficult concepts through analogies. For example, students can ask GenAI to analogize a "randomized controlled trial" to a chef trying out a new recipe to help high school students better understand this research method.

10. Role-playing: simulating conversations with users and audiences

Synopsis. Simulate conversations with possible users and audiences. For example, students could take on the role of an instructor in an online software engineering master's program and engage in a conversation with GenAI to understand the challenges they face in managing their workload.

11. Reinterpretation: seeking a different interpretation

Synopsis. Deepen understanding by prompting different explanations. For example, students can ask GenAI to explain the concept of "carbon sequestration" in different ways, even at a fifth-grade level of understanding.

12. Analytics partners: providing personalized future directions

Synopsis. Provide personalized future directions based on the data collected. For example, a student could provide GenAI with observational data about a five-year-old boy and ask for ideas for toys that would be suitable for making in a makerspace.

13. The Big Idea: Identifying the Heart of the Text

Synopsis. Recognize the core ideas of a text. For example, students can ask GenAI to extract the three most important concepts from a PDF document about the "egg landing" experiment designed by NASA's Mars rover.

14. Designed for Everyone: Exploring Project-Specific Accessibility Recommendations

Synopsis. Explore project-specific accessibility recommendations. For example, students could ask GenAI about accessibility considerations that should be taken into account when creating physical guides for older adults, including font size, font type, and layout design.

15. Visualization: creating visual representations of project ideas

Synopsis. Create visual representations of project ideas. For example, students could ask GenAI to generate a virtual reality game interface that includes magic farmland, different types of crops, and a game interface overlay.

16. A Dream Come True: Creating Images with Limited Existing Reference

Synopsis. Create images that didn't exist before. For example, students can use Firefly to generate an illustration for a children's book about climate futures, showing a city that mixes indoor and outdoor spaces.

17. Survey design: development of a preliminary draft questionnaire

Synopsis. Develop a draft of the initial questionnaire. For example, students could ask GenAI to help design a survey to group their projects based on their reading habits.

18. Base layer: remixing music, art and other media samples

Synopsis. Remix music, art, and other media samples. For example, a student could ask GenAI to generate a one-minute instrumental track combining classical and electronic elements for a video game about unraveling a mystery in an abandoned library.

19. Case composition: creating customized case studies for specific audiences

Synopsis. Create customized case studies for specific audiences. For example, a student could ask GenAI to write a case study about a student who struggles with reading and highlight how teachers can support the student's reading process.

20. Telling Your Story: Generating a Project Narrator

Synopsis. Generate project narration. For example, students could upload voice samples of team members and use GenAI to generate narration for a post on an urban design blog.

21. Design stylist: provides design advice on fonts, colors and themes

Synopsis. Provide design suggestions for fonts, colors, and themes. For example, students can ask GenAI for advice on designing a friendly and colorful application for Gen Z users, including font combinations and color palettes.

22. Name Generator: Suggested Project Title

Synopsis. Suggest titles for the project. For example, students could ask GenAI to provide 10 title options for a children's board game about different ecosystems.

23. Multi-input: using text and image prompts

Synopsis. Use text and image prompts. For example, students could provide GenAI with text prompts and style and color reference images to generate a poster about a summer backpacking and hiking camp for teens.

24. Code Craftsmen: assisting in the creation of code snippets

Synopsis. Assist in creating code snippets. For example, a student could ask GenAI to provide the C# code snippet in Unity to enable an object to follow the mouse.

25. Decoding Your Code: Interactive Code Interpretation

Synopsis. Provide interactive code explanations. For example, students can enter code into GenAI and ask for an explanation of what each line of code does and is intended to do.

26. Debugging assistant: recognizes errors, interprets error messages and provides alternatives

Synopsis. Recognize errors, explain error messages, and provide alternatives. For example, a student could provide GenAI with an error message in Python code and ask for an explanation of the cause of the error and how to fix it.

27. Formula Finder: Find Spreadsheet Functions for Data Analysis

Synopsis. Find spreadsheet functions for data analysis. For example, students can ask GenAI about functions that count the number of times a particular word appears in a row in Excel.

28. Friendly Translator: Learn colloquial phrases and slang in other languages

Synopsis. Learn colloquial phrases and slang from other languages. For example, students can ask GenAI for some informal Hindi phrases to thank their program partners.

29. Writer's choice: selective incorporation of language editing

Synopsis. Selectively incorporate language editing. For example, a student can enter a draft into GenAI and request changes to the paragraph to make it clearer, then selectively accept or reject suggestions as needed.

30. Review, Reflection, Revision: reflecting on suggestions for revising written work

Synopsis. Reflect on suggested changes to the written work. For example, students could enter an executive summary of their project proposal into GenAI and ask for feedback on strengths, weaknesses, transitions, and clarity.

31. Ask Anything: Ask Bold Questions Often

Synopsis. Frequently ask bold questions. For example, students can ask GenAI questions and seek advice about the social responsibility and authenticity risks of mental health and wellness apps.

32. Devil's advocate: providing multiple perspectives to support reflection

Synopsis. Provide multiple viewpoints to support reflection. For example, students could ask GenAI to play devil's advocate and criticize the clues in the movie's plot.

33. What? So what? Now what? : reframing the question and supporting reflection

Synopsis. Redefine the problem and support reflection. For example, students can discuss with GenAI how to improve the display of "error" messages in games and explore other ways to promote growth mindset.