back






Inquiry AI: Reintroducing Friction
Ethnography, Systems Thinking, Prototyping, Social Impact, User Experience
Course: Senior Capstone — Parsons School of Design

compendium 

inquiry-ai-puce.vercel.app



A research and design project exploring what happens to critical thinking when AI does the thinking for us. Inquiry AI started with a simple observation: students know AI is making them worse at thinking — and use it anyway. The challenge was to understand why, and whether better tools could change that. Through six months of ethnographic research and interviews with students, three patterns emerged: the Competitive Trap, Cognitive Migration, and the Awareness Paradox. The result was a Socratic learning tool tested with 45 high school students — rigorous enough to be research, tangible enough to be real. The biggest finding wasn't about the tool. It was about the system: good tools can't overcome bad systems.



critical thinking
ai in education
behavioral research
prototype testing
socratic design
systems thinking



role: solo — 
research design
interviews
prototype development
user testing

duration: one semester

methods: 
ethnographic interviews
behavioral research
prototype design
structured user testing



















Framing the Challenge:

AI tools are now embedded in how students work. The dominant institutional responses — banning AI, adding AI literacy curriculum, updating honor codes — all treat AI as an external threat to manage. This project started from a different observation: students already know AI is affecting their thinking. They talk about it openly. They use it anyway. The problem wasn't awareness. It was something structural underneath the behavior.














                       









The Strategic Question?



How might we redesign AI so it functions as a peer rather than an answer machine — making deep thinking competitively advantageous, not a competitive sacrifice?



The project had one clear constraint from the start: banning AI wasn't a viable solution. Any intervention that required students to opt out of a tool giving them competitive advantage in a high-pressure academic system was going to fail before it started. The question wasn't how to remove AI from the equation — it was how to change the relationship students had with their own thinking while AI was still in the room.







                                                                                                                                         





The project started broad — AI, cognitive atrophy, instant gratification. The pull toward cognitive atrophy was deliberate: it was unfamiliar territory, which felt worth the risk. Early interviews at Pace, Cardozo, and The New School sharpened the question from a general interest into something specific: what will happen to humanity if we automate critical thinking with AI? By week five that had evolved further — not just what happens, but what does society look like in 10, 20, 50 years if we continue down this path?


The deep interview phase produced three insights that drove everything after.



Cognitive Migration came first.

Students weren't just using AI for convenience — they were actively offloading cognitive tasks they used to do themselves. Writing, structuring arguments, generating ideas. The thinking was migrating out of their heads and into the tool.

The Competitive Trap followed. 
Students understood the tradeoff. They knew AI was weakening certain skills. But in a system where peers were using it and grades depended on output, opting out felt irrational. The trap wasn't ignorance — it was incentive structure.

The Awareness Paradox was the most unexpected. 
Students who were most aware of AI's effect on their thinking were also the heaviest users. Awareness wasn't protective. It created a kind of resigned acceptance — they saw the problem clearly and felt unable to change it anyway.

Together these three insights pointed to the same conclusion: this wasn't a tool problem. It was a systems problem.



Research and Insight Development






 




























Strategic Framework

Three intervention directions emerged from synthesis: redesigning AI to function as a peer rather than an answer machine, redesigning educational assessment to reward struggle and process, or building awareness tools that helped students track their own cognitive offloading.

Using an opportunity matrix mapping impact against feasibility within a semester timeframe, I landed on AI redesign — specifically Socratic AI. Educational redesign required institutional buy-in that wasn't available. Awareness tools addressed the symptom, not the behavior. 

Socratic AI was the highest-impact option I could actually build and test.

The core logic: an AI that refuses to do cognitive labor for students. Instead of giving answers, it asks questions. Instead of completing thoughts, it creates productive friction — the kind that Andrew Shea's creative friction framework identifies as essential for deep learning. Remove the friction and you remove the mechanism for development.















Implementation

The prototype had two modes. Peer Mode simulated a peer challenging your thinking — conversational, lower stakes, back and forth. Mission Mode framed critical thinking as a challenge to complete — structured, goal-oriented, with a clear sense of progression.

The prototype was built in HTML/CSS/JavaScript, deployed via Vercel with Google Sheets backend tracking. First-round testing with people I knew surfaced UI issues and informed refinements before moving to the high school cohort. Mission Mode was made more accessible based on early feedback — and became the mode students consistently gravitated toward and asked for by name.

The deployment strategy recognized the core adoption barrier: students won't voluntarily choose friction. The solution wasn't persuasion — it was infrastructure. The roadmap moves from standalone web app shared by teachers in months one through six, to deep LMS integration with Canvas and Blackboard where teachers flag assignments as "Inquiry AI Required." Not a choice. Infrastructure.




























Impact and Learning

Testing with 45 grades 11-12 students across multiple classes produced a clear pattern: high initial resistance, genuine value after use. Students who pushed back — "I wanted the direct answer," "annoyed by repeatedly asking me questions" — were often the same students who said afterward "it made me feel like I was learning and not cheating" and "felt like I was actually using my brain." That gap between resistance and recognition is the whole problem in miniature.

The tool worked. The system around it didn't change. Students who valued what Inquiry AI did for their thinking still wouldn't choose it voluntarily in a context that rewarded output over understanding. That finding reframed the entire project — from tool design to systems intervention. The question isn't about banning AI or teaching responsible use. It's about redesigning the system. Do we optimize for frictionless — or for learning?

What I'd do differently: involve educators and administrators earlier. The adoption barrier was never student motivation — it was incentive structure. Understanding that sooner would have made the deployment strategy the center of the project rather than the conclusion.