Decoded #1: Deterministic vs. Probabilistic

Two words that get thrown around a lot in clinical AI. Here's what they actually mean.

A deterministic system is like a recipe. Same ingredients, same steps, same result every single time. If you put in two cups of flour today and two cups of flour tomorrow, you get the same cake. Predictable. Boring, in the best possible way.

A probabilistic system is more like asking three excellent cooks to make you dinner. They'll all make something delicious, but no two meals will be identical. There's creativity, reasoning, variation, and the occasional surprise.

Most modern AI is probabilistic. Ask a chatbot to draft the same email twice and you'll get two different emails, both reasonable, neither identical. That's also why it sometimes tells you something that sounds confident but isn't true. The same creativity that makes it useful is what makes it unreliable.

In clinical care, you usually want the recipe. When a patient with type 2 diabetes and stage 3 kidney disease shows up, you don't want three different recommendations depending on what mood the AI is in. You want the same answer, grounded in the same guideline, every time.

That's why the rules engine that scores medication recommendations at DisEASE is deterministic. The creativity comes later, in how we explain the recommendation back to the clinician. The decision itself doesn't get to be creative.