CHIPS XR

OVERVIEW

What if students could learn a chip by building it?

CHIPS XR is a mixed-reality learning experience developed at the University of Florida. Students explore semiconductor structures, assemble components and revisit steps without the constraints of a physical lab.

I designed the initial UI and interactions, planned and ran the first usability test, then led the UX direction after that evaluation. I worked with the PM and developers to translate observed problems into clearer interaction cues, learning flows and contextual guidance.

In a later research-team study (n=24), MR+AI showed 21% higher spatial-knowledge scores and 24.17% lower extraneous cognitive load than video. These are condition-level outcomes of the evaluated experience, rather than the isolated effect of my design work.

THE CHALLENGE

The diagrams were flat. The structures were not.

The team’s review of prior research and educational materials pointed to two constraints: students had to mentally reconstruct layered structures from 2D diagrams, and had limited opportunities to practise with real equipment.

We explored whether direct observation and assembly could help students understand spatial relationships, and whether repeatable MR practice could let them revisit difficult steps at their own pace.

FIRST PROTOTYPE

Explore the whole. Examine a part. Build it back.

The first flow started with a complete structure, let students inspect components one at a time, then asked them to reconstruct what they had learned. Invalid placement led to another attempt.

01 / Explore

02 / Examine

03 / Reassemble

The whole gives learners a reference. Isolating a part focuses attention on its relationship with the structure. Reassembly makes that understanding observable through action.

USABILITY TESTING

Easy in the survey. Less clear in practice.

I tested the prototype with 12 engineering students using think-aloud tasks, behavioral observation, and pre- and post-study questionnaires in Qualtrics.

The responses did not tell the whole story. Students still stopped to interpret what could be grabbed, where a part belonged and whether an action had finished. Learning goals became mixed, and looking outside MR for help interrupted the experience.

DESIGN DECISIONS

Make the learning clear, starting with the interaction.

I reframed the observations into three design priorities, then worked with the PM and developers to agree on what to change in the next version.

01 / INTERACTION CLARITY

Show what can be used, what is selected and where it belongs.

Scattered components increased the search area. Ambiguous state cues made students interpret the interaction rules themselves. I brought components into one panel, distinguished selected and used states, and made placement targets and feedback visible in the learning space.

Before

After

A shared panel reduces the area students need to search. Visible component states and placement feedback help them continue without asking the facilitator to explain each action.

02 / LEARNING STRUCTURE

One learning goal at a time.

Exploration, assembly and process explanations had been mixed into one experience. I separated them into Learn, Assemble and Fabricate, so each mode had a clearer goal and expected action.

Learn

Observe components and their layered relationships.

Assemble

Rebuild the structure and check understanding through action.

Fabricate

Use explanations and demonstrations to understand the process.

Learn and Assemble focus on structure. Fabricate focuses on process. Separating those goals reduces the need to work out what should be learned while also learning the controls.

CONSISTENT UI RULES

Make the next interaction feel familiar.

The first study showed why consistent rules mattered. I consolidated recurring UI patterns for component exploration, interaction states and feedback, so new learning content could use the same visual and behavioral language.

EVALUATION

Test the learning value—and keep the limits visible.

The research team evaluated the updated experience with 9 target-course learners through think-aloud sessions and focus groups. The team used MAXQDA thematic analysis to understand how the learning experience and guidance were received.

In a later repeated-measures study, 24 engineering students experienced Video, MR Only and MR+AI conditions. The UX researcher and faculty led this study. The values below compare condition means; they do not isolate any single design change.

The slides also report immersion at 4.65 / 5 for MR+AI and usability at 4.11 / 5 for the interactive conditions. These are the reported study measures, not a standard 0–100 SUS score.

MR showed its clearest value in spatial understanding and exploration. Conceptual and procedural learning did not show a clear advantage over the existing approach, and the short study did not establish long-term learning retention.

Research reference: Cheon et al. (2026), Learning Inside the Chip. CHI ’26. DOI: 10.1145/3772318.3793705

REFLECTION

Watch where people stop. Then ask why.

Running the first study changed how I judge a design. A polished screen and positive questionnaire responses were not enough. I learned to look for hesitation, connect observed behavior with participants’ explanations, and use quantitative results as another part of the evidence.

We did not establish how well learning would be retained over a full course, or which questions and situations benefited most from AI support. Initial guidance also remained important: a useful learning experience was not automatically easy to start.

I would clarify the division of roles: use MR where spatial understanding matters, and AI where contextual explanation helps. In a real course, I would study retention and AI question patterns, strengthen the features with evidence, and reduce those that add little value.