One-way vs. two-way decisions

There’s a thing in machine learning called accuracy, telling you how good an algorithm is.

The lesson you learn quickly is that accuracy can be a terrible measure of success. Take cancer detection, for instance. Most probes for cancer come back clean. Only a very few actually are cancerous.

That means any algorithm that predicts “non-cancer” 100% of the time is pretty accurate, right most of the time.

And yet, it makes an error in 100% of the important cases.

I feel like it’s the same with one-way vs. two-way decisions. Most of us focus on making decisions quickly because most are two-way decisions; we can go back.

And yet, we almost certainly miss most of the one-way decisions that way.

If you want to be serious about decision-making, I think you need to spend a considerable amount of time thinking about whether a decision is more of a one-way vs. more of a two-way decision.