Cognitive Games — Akshat Sharma

Clinical cognitive assessments are built to measure, not to entertain. But for cognitive therapy, engagement is part of the product: the science team needed around 20 minutes of sustained engagement, and the existing assessment flows held attention for 3–4.

I helped transform four scientific tasks into a suite of mobile-first games for people with Mild Cognitive Impairment — preserving the approved assessment logic while adding story, scoring, feedback, and interaction patterns that give people a reason to continue.

The four games: Mystic Jungle Quest, Signal Switch, Space Cards Memory, and Maze Walker — Turning clinical tests into an engaging therapy — the four-game suite.

The 20-minute engagement gap

3–4 min

attention held by the original assessment flows

20 min

engagement required for therapeutic benefit

4 games

designed and prototyped to close the gap

Bar chart showing 3-4 minutes of current engagement vs the 20-minute therapeutic requirement — The gap that made this a strategy problem, not a reskin.

Three constraints shaped every decision

Approved mechanisms — the core task logic was already clinically approved and couldn’t be casually altered
Existing backend — new experience features had to layer on top of what was built
React Native web app — the platform limited complex animation and sprite-based game behavior

My role was to turn those constraints into a design strategy, not a list of compromises.

Three constraint cards: FDA approval, existing backend, web-app platform — Designing within three constraints — FDA approval, the existing backend, and the web-app platform.

Strategy: wrap the science in a story

Keep the scientific task intact; redesign everything around it. Each clinical task became a game concept with its own metaphor:

Clinical task	Game wrapper	Design move
PVT	Mystic Jungle Quest	Abstract stimulus-response becomes collecting treasures, avoiding traps
Flanker	Signal Switch	Directional attention becomes a character-led interaction system
N-Back	Space Cards Memory	Complex working memory simplified through card matching and sliding turns
Maze	Maze Walker	Validated mechanic refreshed with level differentiation and a character world

Diagram mapping each clinical task to its game wrapper — The core strategy: keep the approved task, redesign everything around it.

From abstract colors to treasures and traps

The first PVT approach used simple color stimuli and rounds — too thin, and fast users finished too quickly to hit the therapy target. The breakthrough was a collection/avoidance loop: users weren’t responding to abstract colors anymore, they were collecting treasure and dodging traps in a jungle world. Same response logic, real motivation.

Three-stage PVT progression from color stimuli to the jungle treasure metaphor — PVT's three lives: abstract colors → time-based rounds → treasures and traps.

Mystic Jungle Quest (PVT) — the final collect-and-avoid loop in motion.

Build a system, not just a screen

The Flanker character had to respond across screen sizes without a game engine. A one-off illustration would break; a system would scale. I constructed the character from configurable parts — each body element separate and composable — so the design stayed robust across layouts and stayed buildable for the web-app platform.

Flanker character broken into modular, configurable parts across screen sizes — The Flanker character as a system — every body part a configurable element.

The breakthrough came from watching, not designing

N-Back was the hardest task to communicate. Multiple metaphors failed — lifting cards, stacking objects, prettier visuals that still leaned on text instructions. Observation broke it open: the separate “Match” button was redundant (the card itself was the natural target), and explanations collapsed above N=2. Sliding cards made the metaphor click — a new turn arrives, you compare it with memory, and the action lives on the object itself.

Failed assumption

Observed insight

A better visual would solve comprehension

The interaction model itself was unclear

Text instructions could carry the task

The metaphor and affordance had to do the teaching

Users could start at higher complexity

N=2 had to be understandable first

Three failed N-Back metaphors marked as 'didn't work' — Three attempts that didn't work — lifting cards, stacking objects, prettier-but-still-confusing.

The N-Back breakthrough: card as the interaction target, sliding-turn metaphor — The breakthrough: kill the Match button, start at N=2, let the card be the action.

Space Cards Memory (N-Back) — the sliding-card metaphor in motion.

Maze Walker: respect what’s validated

Not every game needed reinvention. The Maze mechanics were already validated — the design task was differentiation and appeal: clearer levels, a friendlier character world, and a UI refresh that helps users orient without touching the validated core.

Maze Walker character explorations and level differentiation through backgrounds — Maze Walker — character exploration and level differentiation through backgrounds.

Maze Walker — the refreshed, character-led maze in motion.

Testing with two extremes

Rapid informal testing stress-tested clarity against two extreme behaviors: one user who read every instruction before starting, and one who skipped straight to playing. If the game could explain itself to both, it could survive the real world.

Two extreme feedback personas: the 'reads everything' user and the 'no instructions' user — My secret weapons — the 'reads everything' user and the 'no instructions' user.

Outcome

Four prototyped games that preserved the scientific requirements while creating a genuinely more engaging experience layer — aligning science, product, and engineering constraints in one coherent suite.

The final four games as a coherent therapeutic suite — From clinical tools to an engaging therapeutic suite.

What I’d do differently

More research documentation, earlier low-fidelity prototypes. Both are now standard parts of my process — process over pixels.