Computational Making in Public-School Classrooms at Scale
A year-long deployment of computation-based Making in authentic science classrooms and a 340-hour video analysis of how students actually connect science, Making, and code together.
Project Overview
Adding computation to hands-on "Making" is a powerful way to teach science - a 3D-printed model becomes dynamic and interactive when an Arduino and code drive it. But almost nobody had studied whether this works in ordinary public-school classrooms at real scale, where curriculum standards, logistics, and hundreds of students come together. Our program did exactly that, for a full year, across every 5th- and 6th-grade science class in a school.
My role. Within a large (69 members) multi-team program, I led the research strand as first author on two papers: the retrospective Design-Based Research analysis of what it takes to deploy computation-based Making at scale, and a 340-hour video analysis of how students actually make interdisciplinary connections. I owned the research questions, coding schemes, and synthesis into design implications.
The impact. The work delivered the field’s first comprehensive account of at-scale computation-based Making in authentic classrooms: four systemic challenges that any scaling effort must design around, and a video-grounded model of how (and how rarely) students link science, Making, and code. These findings directly shaped the program’s year-two iteration and offer a reusable playbook for schools and ed-tech teams.
The Challenge
Piloting a maker activity with a handful of motivated kids tells you little about running it for 442 students inside a mandated curriculum. Computation raises the difficulty further: students juggle science concepts, physical building, and programming at once, and teachers rarely have all three. The challenge was to make computation-based Making survivable and effective in the constraints of real classrooms - and to understand where the intended learning connections actually form.
Core questions we needed to answer
- What does it actually take to deploy computation-based Making across a whole grade for a year?
- What challenges impact the sustained, curriculum-aligned use of computation-based Making?
- What types of connections do students make across science, Making, and computation?
- In what contexts do those interdisciplinary connections emerge?
Methodology
Because the goal was to improve a real, evolving intervention rather than test a fixed hypothesis, we grounded the program in Design-Based Research - iterative cycles of design, deployment, and reflection in the live setting. Different components (Maker kits, mentor corps, organization) iterated on different clocks. For the learning question, I paired this with fine-grained video interaction analysis.
Design-Based Research at program scale
Why Authentic classrooms are too dynamic for a controlled trial; DBR is built for improving practice in messy, real-world settings through iteration and practitioner collaboration.
How Coordinated iterative cycles across specialized sub-teams (fabrication, curriculum, mentors, a custom block-based programming interface, and research), refining curriculum-aligned maker kits and week-long lesson plans between interventions and capturing organizational lessons for the next year.
Curriculum-aligned activity & lesson design
Why To be adoptable, activities had to map onto what teachers were already required to teach.
How Co-designed activities with teachers around existing science units (e.g., Electricity in Circuits, Movement of the Sun), each pairing a physical build, block-based code, and target science concepts into structured multi-day lesson plans.
340-hour video interaction analysis
Why Self-report can’t reveal the micro-moments where a student links a science idea to a wire or a line of code. Video captures interdisciplinary connections as they happen.
How Analyzed ~340 hours of classroom video, building a Knowledge×Application linkage model across science, Making, and programming; two coders identified 68 connection excerpts and coded both the linkage type and the context that produced it (Cohen’s κ = 0.72 and 0.80 - substantial agreement).
Key Insights & Artifacts
Scaling surfaced four systemic challenges - the design brief for at-scale Making.
Across the year, four challenges recurred: striking a workable balance between science, Making, and computational thinking; making the connections between those disciplines explicit to students; the sheer organizational and operational overhead (kit logistics, device reliability, mentor coordination); and delivering the designed lesson plans with fidelity in unpredictable classrooms.
Decision These four become the requirements any school or ed-tech team must budget for up front - the program’s value is as much in operational design (logistics, mentor training, kit robustness) as in the learning activity itself.
Students’ connections were lopsided - computation stayed invisible unless made tangible.
The dominant connection by far was science-knowledge to Making-application (≈68% of coded moments): students readily used science ideas to guide hands-on building. Programming-related links were rare and fragile, surfacing mainly when computation was anchored in something observable (an LED turning on, a sensor reading) or explicitly prompted by a mentor.
Decision If computation isn’t designed into visible, tangible touchpoints, it recedes into abstraction and the intended learning connection never forms - a direct implication for activity design and for where mentors should intervene.
Hands-on, embodied activity is what actually produces the connections.
Coding the contexts behind connections showed they were driven by physical, embodied engagement: cause-and-effect experimentation (the single largest context), tinkering, and embodied metaphors (students reasoning about their circuit as a model of a concept) - often catalyzed by a mentor’s well-timed question.
Decision Design should deliberately engineer these moments - tangible cause-and-effect, room to tinker, metaphor prompts, and trained mentors - rather than assume connections emerge on their own; without scaffolds, opportunities are easily lost.
Impact & Learnings
How the team applied the findings
The retrospective lessons and the video-grounded connection model fed straight into planning the program’s second year - reshaping kit design, mentor training, and where computation is surfaced in each activity.
- First comprehensive, documented account of at-scale computation-based Making in authentic public-school classrooms.
- Four systemic scaling challenges named and evidenced - a reusable checklist for schools and ed-tech teams.
- A validated interdisciplinary-linkage model (κ = 0.72-0.80) plus seven contexts that produce connections.
- Concrete design implications: make computation tangible, engineer cause-and-effect moments, and equip mentors to prompt cross-discipline talk.
What I learned
At scale, the intervention is only half the product - logistics, device reliability, and mentor capability determine whether the learning design ever reaches students intact. I also learned how easily the "computational" half of computational Making disappears: unless code is tied to something students can see and touch, they simply route around it.
What I'd do differently
I’d surface computation earlier and more visibly in each activity, front-load mentor training specifically on facilitating cross-discipline talk (since mentor prompts were a key trigger for the rarest connections), and instrument activities for learning outcomes so we could tie specific design choices to measured understanding, not just observed connections.