Navigating Student-Directed Projects with GenAI: Practical Strategies from Harvard’s Field Study
I came across a report exploring a crucial question:
What happens when students drive their own learning, with GenAI riding shotgun?
Through interviews with 27 Learning Design students and 7 faculty, a group of students and educators at the Harvard Graduate School of Education captured how generative AI tools like ChatGPT, Claude, and Perplexity are reshaping how students brainstorm, organize, revise, and reflect during long-term, open-ended projects.
While the context was graduate-level learning, the findings offer constructive, cross-cutting insights for anyone guiding student-driven work, be it high school capstones, college seminars, or independent studies.
I’ll outline some of the most practical strategies educators can apply with their students.
Phase 1: Getting Started with AI—Scaffolding the Uncertain
1. A Starting Point: Orienting the Learner
One of the simplest and most valuable uses of GenAI is helping students gather basic information and orient themselves at the start of a project. Instead of endlessly Googling or freezing up at a blank document, students used AI tools to ask exploratory questions: What is a solar eclipse? What is “maker education”? They weren’t searching for end-all answers; they were looking for a spark that might create an intellectual opportunity.
When students explore a new domain, even small gains in understanding can dramatically reduce deflating barriers. As one student explained, “I think [AI] can be a second brain for you if you’re feeling a little stuck. You can ask it to help you get started, but it can’t do everything for you. You’re still the pilot and it’s just an assistant.”
2. Structure the Chaos: Planning and Pacing
Plenty of students full of energy and ambition at the start of a project quickly become overwhelmed and lose steam weeks later. (I observed this frequently over 15 years of overseeing independent high school projects.) Fortunately, GenAI excels at breaking complex goals into manageable pieces.
Students used ChatGPT to generate step-by-step plans, timelines, and week-by-week deliverables. For instance, a student creating a museum exhibit for middle school girls asked AI to generate a project calendar complete with objectives and tasks.
The takeaway here is that the AI wasn’t replacing planning, it was supporting executive function. The AI was acting like an external project manager, nudging students toward sequence and accountability.
3. Define It: Clarifying Constructs
When these Harvard students were exploring big concepts, such as “student engagement” or “agency,” AI helped them make sense of abstract ideas. One student asked, What are the features of student agency? and received a list of 20 characteristics. The student didn’t take it as an indisputable truth, but rather as a springboard to synthesize, critique, and combine ideas.
As one student said:
ChatGPT is more forgiving about me waking it up and asking questions. If I want to ask a professor something, I want to make sure I don’t sound stupid. All of those social constraints holding me back from asking the question the way I want to ask it are gone. ChatGPT has no constraints. You just ask whatever you think.
4. Scope and Scale: Guardrails for Big Ideas
It’s easy for student-directed projects to balloon into unmanageable beasts. (I’ve seen a few student tears shed over this.) A key strategy shared in the guide was using GenAI to set project constraints. One student asked: This is starting to feel too big. How can I scale it back while preserving [the core elements]?
AI helped students prioritize essential components while eliminating excess—a skill many adults (hand raised) still struggle with. It’s not that GenAI makes these decisions for students, but it provides a perspective and potential counterbalance when ambition meets reality.
In partnership with Essay Grader
Essay Grader is a straightforward, powerful tool that uses AI to take the load off your shoulders and will completely change the way you approach grading!
Every feature puts teachers first, with a focus on simplifying repetitive tasks like grading to improve productivity - all the while letting teachers be fully in control.
Streamline grading, provide detailed, actionable feedback to your students, and maintain complete control over your assessments - all in one intuitive interface.
Reduce grading time by 80%. Grade your entire class's essays in 2 minutes or less and deliver high-quality, specific feedback to your students.
Ensure fair essay grading by removing bias and aligning with state standards like Texas STAAR, Florida BEST and California CCSS.
Join Essay Grader and its ever-expanding network of 60,000+ educators, and you’ll connect with others who share your passion and commitment.
Phase 2: Iteration and Feedback—Making the Work Better
5. Roleplay and Persona Building
Whether building an app, writing a story, or developing a learning experience, the students in this study often struggled to see beyond their own point of view. GenAI allowed them to simulate other voices: What would a faculty member say? What would an 8th grader care about?
Conversing with personas, students developed a more nuanced understanding and clearer user-centered designs. While the AI’s suggestions weren’t always accurate, even occasional mispeak generated discussion and refinement.
6. Conceptual Parallels: Clarifying the Abstract
As we know, AI is remarkably good at drawing analogies. And students took full advantage. When struggling with technical or theoretical texts, they asked for comparisons and metaphors: Explain carbon sequestration like I’m five. Use a kitchen metaphor.
They weren’t trying to dumb down content; they were putting abstract concepts in mental models. What’s notable is that the students didn’t just consume the analogies—they evaluated them, adapted them, and created their own. In other words, they created a scaffold for deeper understanding.
7. Dream into Existence: Visualizing the Impossible
When students imagined things that didn’t exist—a decolonized math textbook, a climate-resilient city—they used AI image tools like Adobe Firefly to bring their vision to life. For non-illustrators or first-time designers, this was a way of prototyping ideas.
As one student explained:
I needed to dream up something that’s never existed before. My idea is to create a math textbook that decolonized mathematical history. It places mathematics in its indigenous history, and those images don’t exist. I used Firefly, because I’m not a great artist. That was reallyhelpful in showing people what things could look like that have never existed before.
8. Survey Design and Data Collection
Several students used GenAI to design early versions of surveys or data collection tools for their projects. One student, building a reading-based grouping system, asked ChatGPT: How should I structure a survey to sort students by genre preference and reading habits?
The AI offered a Likert-scale-based structure, suggestions for genres, and tips for clarity and analysis. As with many strategies in the guide, the value came not from outsourcing thinking, but from having a capable assistant who could get the process moving faster.
Phase 3: Expression and Refinement—Communicating Ideas Clearly
9. Review, Reflect, Revise
For many students, GenAI became a revision partner—not just for grammar, but for structure, clarity, and logic. Students asked questions like: What are the strengths and weaknesses of this executive summary? Are the transitions smooth?
The AI responses weren’t always perfect, but they were consistent, nonjudgmental, and available 24/7. GenAI became a tool for improving fluency without shame.
10. Devil’s Advocate and Multiple Perspectives
One clever strategy was asking GenAI to “play devil’s advocate.” This prompted AI to offer counterarguments, gaps in reasoning, or alternative viewpoints. Whether students were writing a film plot, designing an app, or defending a thesis, this approach sharpened their ideas and surfaced blind spots.
In short, GenAI became a low-stakes peer reviewer—quick to disagree, but not to judge.
11. What? So What? Now What?
This strategy, near the end of the guide, was using GenAI to support reflective thinking. Students described moments when they were stuck and unsure how to even define their problem. GenAI helped them rephrase, clarify, and interrogate their assumptions.
As one student explained”:
“Sometimes you get stuck with your problem. You don’t understand your problem and you really need to rethink what you’re doing. GenAI challenged me to rethink and define and rephrase my problem until I actually understand it. And I learned a lot about metacognition here. You talk to GenAI, and it gives you ideas about what your problem is.
Nuance, not cognitive offloading
What struck me most about the report was the nuance of the thinking. Students weren’t outsourcing their projects to AI. They were using it thoughtfully, critically, and creatively. They experimented. They iterated. They questioned what it meant to “get help” in a world where assistance can be instantaneous.
The guide reminds us that GenAI shouldn’t flatten the learning process. Instead, if used well, it should enrich it. That said, the key is design: clear goals, reflection, and scaffolds that center learning rather than completion.
As one student put it:
“Do you just want to get the job done, or do you want to learn?”
Indeed, one of the guide’s most powerful strategies, “What? So What? Now What?”, positions GenAI as a reflective tool that helps students reframe their understanding of a problem. In this strategy, students use GenAI not to find solutions, but to interrogate the assumptions behind their own questions, prompting self-examination and deeper understanding.
We all know some students will use AI to get the job done. But the report explicitly and constructively outlines how GenAI can enhance critical thinking by provoking metacognition, generating alternative perspectives, and encouraging iteration.
AI in Edu: News, Tools & Views is sent to all subscribers on Fridays.
Find more AI tools, views, and how-to’s at tomdaccord.com
Find my book AI Tools & Uses: A Practical Guide for Teachers at Amazon
*Learn more about Essay Grader