Malaria control as a tower-defense game

Many simulation models of malaria have the same challenge: rich detailed dynamics under the hood, but simplistic low-information-content outputs like time-series prevalence curves. I've often wondered how one can take inspiration from video games and provide a richer and more visual platform to develop intuition about dynamics, learn about tradeoffs between strategies, etc.

Over the summer I had a bit of down time to see what that might look like, leveraging notably better capabilities of Gemini in producing game-ready visual assets from prompts, and inspiration from Princeton Tower Defense by Kevin B. Liu (MIT-licensed).

Here's links to where you can play in your own browser. No install — it runs entirely client-side:

• Village-level level that might be refined towards scenario challenges (link)

• Debug tutorial scenario that isolates net-vs-baseline effects and dependence on parameters (link)

Below are a few screen captures of where things landed. It's a small, browser-based strategy game about stopping malaria in a village. You place real interventions — bednets, indoor spraying (IRS), larviciding, clinics and health posts — against a seasonal, mosquito-borne epidemic, and try to keep the community's infection level below a failure threshold across the transmission season.

The twist: under the cartoon-ish presentation is an honest disease-transmission model (individual people with acquired immunity, two mosquito species with realistic breeding and biting behaviour, a proper incubation period, etc.).

Screenshot of daytime view of village, larval habitats, agricultural fields, and health facility. Clinical cases in the population are highlighted with red circles. Health-seeking represented as movement to and from clinic.

https://video.wixstatic.com/video/101ec8_64749d7161d24f82a57e345f050ace59/720p/mp4/file.mp4

Video showing gameplay of placing interventions like a health facility in relation to its catchment.

https://video.wixstatic.com/video/101ec8_971c069c62134a578f34d395b9a54082/720p/mp4/file.mp4

Video of nighttime mode in a debug tutorial setup of two compounds, one with bednets only, the other with IRS only. Mosquito age (cyan) and infectiousness (yellow) represented in swarm colors. Human infectiousness represented by plumes of yellow. Infection state of individuals by their color. ITN protection by the hemispheric dome, IRS by the green perimeter, each with its own corresponding blocking/killing animation.

I was mostly focusing on the visuals and the animations in this prototype. I think they turned out pretty interesting, although clearly I've been trying to cram too many variables and dynamics into the visualization -- more than can really be digested at once.

A few areas that I'm interested in thinking about in the future:

1. Finding the sweet spot for realism of dynamics under the hood for this style of simulation, especially where one wants individual representation vs. individual icons on screen are representative of cohort simulation state

2. Visualization tradeoffs in representing each day/night cycle as 5 or 10 or 30 days in the annual calendar.

3. Extending some of the debug scenarios into more purposeful educational tools

4. Refactoring the gameplay to something that's actually fun to play, not just a simulation -- whether that borrows elements more from RTS games or tower-defense games.