The Computer Science of Human Decisions

Looks like great contribution. In looking for an earlier related work[1] that was well-known in its day, I came across this[2] interesting article with an updated perspective.

[1] https://global.oup.com/academic/product/simple-heuristics-th...

[2] https://behavioralscientist.org/simple-heuristics-that-make-...

One of the most interesting parts of this book for me was the section on exponential back-off, both for the origin of the technique and examples in the real world such as for handling criminal sentences.

I've read this book and found it worthwhile.

A couple of the "algorithms" that I find were worth keeping in mind are:

1. Marriage problem (aka Secretary, suitor, best choice): if you're given options that you can compare but do not have an objective value of worth and you have a bound # of candidates, only look for the first 37% of options. Then pick the next that beats the best-so-far.

I like this one because it shows that for some problems there is no wins-all solution. Sometimes the suitors just get worst over time (real-life dating amirite?). Sometimes the options only get better (computer hardware? Cloud solutions?) but you have to pick at some point.

https://en.wikipedia.org/wiki/Secretary_problem

2. Deadlines make prioritization and scheduling nigh impossible. You can have your priorities all in line and think you can just work from most to least important, but if something is due, like Taxes or putting the trash bins out, those get jammed in.

This is partly why I loathe how deadlines are chosen in the corporate office. Management found that setting deadlines gets things done by a date, so they use it for everything, which throws prioritization of work completely out of wack. Missed deadlines are easy to track and punish and put what's trackable over (potentially) what's important.

3. If you want to know how much longer something will go on or last, and you have no way to estimate, you can assume you're in the middle. Kind of like the law of averages. No one wants to think they're average, but we most likely are (or at least within a std-dev or two.

I find this helps when deciding on, say, what language or framework to learn or which car to buy. It helps to fight against hype. Will X be around for 10 more years? Well, it's only been around for 3, so probably not, but A has been around for 50 years, so...

It also helps to know that whatever latest trend or fad will likely be gone as quickly as it came.

4. I can't recall the formula now, but I know that there's a formula for how long you can expect to wait in a line. It also works for anything with a queue, a process time, and a capacity. It helps when people say "Look how long the line is!" Instead, take a minute and see if the line ever moves, not how long it is.

None of these algorithms completely fit real-life examples, but they just give small tools and tricks to think about things. They're not meant to be used blindly, but just as a starting point to help when there seems to be no starting point or too many options. And hey, sometimes there's just not a "best" and it sometimes helps knowing that too!