Is doing science all about precise thinking?
Most people think doing good science means being rigorous, precise, and analytical. I think they’re only half right.
Lately, I’ve been reading Thinking, Fast and Slow by Daniel Kahneman.1 His core idea is that the human mind operates using two distinct systems:
- System 1: Fast, automatic, intuitive, and emotional.
- System 2: Slow, deliberate, logical, and analytical.
Science looks like pure System 2: controlled experiments, statistical tests, and peer review. That is what gets published, and that is what we teach in universities. But is precision the whole story?
Good science requires novel and bold hypotheses. Where do those come from? Usually, not from careful analysis. They emerge from System 1: hunches, half-baked intuitions, and “what if” moments that don’t yet make sense. Too much System 2 too early kills ideas before they grow. The best hypotheses often sound crazy at first. You cannot always reason your way to a breakthrough; sometimes, you have to let yourself be “imprecise” enough to see a pattern that logic hasn’t caught up to yet.
Of course, imprecision alone isn’t science—it’s just daydreaming. At some point, you must switch. You have to ask: “What would prove this wrong?” That is the principle of falsifiability.2 Doing good science means constantly balancing these two modes. System 1 generates; System 2 verifies. Then, the results feed back into System 1 for the next intuition. The cycle never stops.
The hardest part of research isn’t the analysis itself; it’s the timing. I’ve killed good ideas by demanding rigor too early, and I’ve chased bad ones for too long by avoiding the “System 2” reality check. In my experience, a few questions help guide the switch:
- Too confident without evidence? Switch to System 2. Bring in the math.
- Analyzing for days with no progress? Switch to System 1. Go for a walk.
- Does the idea excite you? Let it breathe. Don’t kill it with a p-value before it has a chance to be articulated.
— Mohit The ideas in this post are my own. Generative AI was used to assist with drafting and editing.
References