Case study · UCLA Psychology · Insight Learning Technology
Perceptual-Adaptive Learning Modules (PALMs): from cognitive science to product PoC
Research-grade perceptual training does not survive contact with learners if the surface is illegible, touch-hostile, or opaque about progress. This write-up is the learner-facing slice of that story—what we fixed in product form, and how it connects to educator-scale delivery (including the no-code path for clinical educators).
tags: POC · Productization · MVP · PoV · GTM · audience: medical students & residents · platform: responsive web app · initial build: 2014 · scale signal: 200+ modules · 1k+ med students (lab program context)
My role
I joined as the first and only design hire on a team that already had the core technology and development framework in place. My job was to partner with the lead domain expert to materialize her vision while working closely with engineers to redesign interfaces and clear usability debt—without pretending the underlying pedagogical model was negotiable.
In parallel, my career context spans UCLA Psychology (programmer analyst work adjacent to the PALMs research line) and Insight Learning Technology (technical product management and design leadership across enterprise education programs). This case study centers the learner-facing PoC trajectory; the companion write-up covers the no-code authoring productization path aimed at clinical educators.
Program context
Perceptual-Adaptive Learning Modules (PALMs) are web- based learning aids grounded in cognitive science. They target capabilities like pattern recognition that traditional didactic instruction often under-trains—especially in domains where “knowing the facts” is not the same as “seeing the right thing fast enough.”
For the company, this effort was also a pivot instrument: a pilot product to move from K–12 math into medical education, with an explicit goal to ship the first medical module. That framing mattered: it forced the team to treat PALMs as a product surface, not only a research artifact—while still respecting constraints inherited from years of prior iteration on the concepts.
Clinical-training context
Clinical training is its own product problem, but it shares a failure mode with other high-stakes UIs: if the interface lies—or merely obscures—people stop trusting the loop. In PALMs, the loop is perception → corrective feedback → retention. A pretty deck does not teach ECG morphology; repeated, well-structured trials with honest feedback cues do.
The honest product read is that many “adaptive learning” products are really content navigation with analytics garnish. PALMs sit closer to perceptual learning: the bar is whether learners build stable, transferable discrimination skills—not whether the LMS checkbox turns green.
Use-case arc we sequenced
When I landed, the science and scaffolding were real; the learner-facing layer was not yet carrying its weight. We sequenced work the way you would in ops product: stabilize the highest-friction interactions first, then tighten feedback semantics, then expand guidance—without rewriting the engine on a whim.
The concrete problem stack included sub-optimal legibility, poor touch-screen usability, ambiguous progress indicators, lack of onboarding, a dated visual language, content-related anomalies, and a learner experience that read as discouraging rather than instructive. Under the hood, years of iteration had left poor practices and outdated libraries in the framework—so even good design intent paid a tax on implementation.
A capability ladder
1. Standardization (components, patterns, dev velocity)
To streamline building PALMs and improve the learner experience, I introduced new components that paired visible UI patterns with consistent behind-the-scenes behavior—so engineers could ship iterations without re-learning a one-off layout every time.
2. Enhanced feedback cues
Perceptual training lives or dies on feedback. I pushed cues to be universally usable, understandable, and fast to parse—so the learner’s attention stays on the discrimination task, not on decoding the UI’s mood.
3. Progressive disclosure and wayfinding
I revised the general flow and inserted dedicated nodes that created real estate for timely hand-holding and way-finding. Those inserts host static and dynamic messages that help learners build an accurate mental model of where they are in the journey—without front-loading paragraphs nobody reads.
4. Messaging and positioning (explain the “why”)
PALMs are unintuitive if you treat them like a textbook with buttons. To communicate what was happening under the hood, I co-scripted and helped create an explainer video aimed at novices—translating technical behavior into a credible, approachable narrative.
PALMs PoC: capabilities snapshot
The following is a capabilities snapshot of the learner- facing PoC as we operated it—meant to separate “what the module runtime was trying to do” from the parallel educator tooling track.
Learner-facing intent
- Web-based modules emphasizing pattern recognition and related perceptual skills—not generic page-turning.
- Responsive delivery: responsive web app assumptions for real devices in the wild (not only lab desktops).
- Continuous iteration after the initial development window— 2014 as the starting point, with ongoing optimization and subject-specific variants over time.
Team shape (representative)
- Domain experts ×3, project manager ×1, engineers ×3, designer ×1 (me).
Evidence of effect (external, not a claim of my PM metrics)
PALMs have been associated—through published research and conference presentations—with improved understanding and automaticity in complex domains, including maintenance of skills over longer horizons for some clinical recognition tasks. The references below are the right place to ground those claims; I am not substituting marketing superlatives for citations.
What shipped and traction
On the learner side, we shipped a materially improved module experience: standardized components, clearer feedback semantics, better onboarding and wayfinding, and clearer external messaging—while wrestling honestly with legacy framework constraints.
- Program scale (UCLA / PALMs research context): on the order of 200+ modules and 1k+ medical students in the adaptive-curricula line I was adjacent to as part of the broader PALMs effort—this is a scale signal from the lab program, not a vanity metric I am claiming personal ownership of.
- Commercial / field trajectory (product narrative): the perceptual-adaptive learning line is part of a broader story where no-code authoring for clinical educators reached adoption, repeat contracts, and revenue traction—the exact business mechanics live primarily in the authoring platform case study, but the outcomes are inseparable in practice: modules do not propagate if educators cannot author them.
References
- Krasne S, Stevens CD, Kellman PJ, Niemann JT. Mastering Electrocardiogram Interpretation Skills Through a Perceptual and Adaptive Learning Module. AEM Educ Train. 2020;5(2):e10454. Published 2020 May 5. doi:10.1002/aet2.10454
- Romito BT, Krasne S, Kellman PJ, Dhillon A. The impact of a perceptual and adaptive learning module on transoesophageal echocardiography interpretation by anesthesiology residents. Br J Anaesth. 2016;117(4):477-481. doi:10.1093/bja/aew295
- Rimoin L, Altieri L, Craft N, Krasne S, Kellman PJ. Training pattern recognition of skin lesion morphology, configuration and distribution. Journal of the American Academy of Dermatology.
- Krasne S, Hillman JD, Kellman PJ, Drake TA. Applying perceptual and adaptive learning techniques for teaching introductory histopathology. J Pathol Inform. 2013;4:34. Published 2013 Dec 31. doi:10.4103/2153-3539.123991
Closing
PALMs are a reminder that “research → product” is not a one-time handoff—it is a negotiation between an evidence base and the brutal clarity of a learner UI. The science can be sound while the experience still fails; my work here was to close that gap enough for the PoC to function as a credible bridge into medical training—and to stay coherent with the educator-facing systems that actually move content at scale.