Algorithms to Live By: The Computer Science of Human Decisions

The summary

Most of life’s hardest decisions aren’t hard because you’re bad at thinking. They’re hard because you’re solving problems that have no perfect answer: too many options, too little information, too little time. Computer scientists have spent decades on exactly these problems, and the strategies they worked out for machines turn out to be good advice for people. An algorithm is just a finite sequence of steps for solving a problem. Following a recipe is one, and so is making a pros-and-cons list, only less precise. The trick is knowing which algorithm fits which decision.

Know when to stop looking

When you evaluate options one at a time—apartments, job candidates, people to date—the agony is committing before you’ve seen enough, or waiting so long the good ones are gone. The optimal stopping rule settles it: look at the first 37% of your options without choosing any, just to learn what “good” looks like, then take the next one that beats everything you’ve seen so far. You won’t always land the single best option, but no strategy does better on average. The point is to stop agonizing and let a rule carry the decision.

Explore, then exploit

The multi-armed bandit problem—named for a gambler facing a row of slot machines—asks whether to keep pulling the arm that’s paid off or gamble on an untested one. The Upper Confidence Bound approach says start with the option that looks best, track what actually happens, and move on when reality falls short of what you expected. Medicine runs a version of this with adaptive clinical trials, adjusting the study as data comes in rather than following a fixed plan to the bitter end. The broader lesson is about timing. When you’re new to a city or early in life, explore: try the untested restaurant, meet new people. When your time is short or you’re passing through once, exploit what you already know is good.

Keep what you use within reach

Computers store the data they touch most often in a small, fast cache near the processor and dump the rest in slow, cheap storage. Your desk, your closet, and your brain all work better on the same principle. Keep what you use frequently and recently close at hand; when space runs out, evict whatever you won’t need for the longest time. This is why reading your notes right before bed loads them into easy reach for a morning exam, and why forgetting a name isn’t a character flaw but, as the authors put it, a “cache miss.” One caveat on tidiness: “Sorting something that you will never search is a complete waste; searching something you never sorted is merely inefficient.” Don’t organize what you’ll never look for.

Do the urgent thing, and only that thing

For deadlines, the Earliest Due Date rule is hard to beat: work on whatever’s due soonest. If you can’t finish everything, Moore’s Algorithm minimizes how many tasks come in late—follow earliest due date, then drop the single longest task when you fall behind. Better to blow one big deadline than three small ones. Watch for priority inversion, the trap where a trivial task blocks an important one, like answering email while the report your boss needs sits untouched. And resist the urge to multitask. Every switch forces your mind to reload its working memory, so the fastest route through a pile of work is one task at a time, finished before you move on.

Good enough beats perfect

The most complex model isn’t the best one. Pile in enough variables and you get overfitting: a solution so tuned to the data you have that it fails on anything new. Real decisions are the same—optimize too hard for one situation and you’ll be wrong about the next. So accept effective over ideal. “Don’t always consider all your options… Make a mess on occasion. Travel light. Let things wait. Trust your instincts and don’t think too long. Relax. Toss a coin.” This extends to how you treat people, through what the authors call computational kindness. Saying “Oh, I’m flexible, what do you want to do tonight?” feels generous, but it hands someone else the whole problem and forces them to guess your preferences, one of the hardest things a mind can do. Just say what you want.

The bottom line

The deepest idea here is that perfect information is a fantasy, and simple rules of thumb often beat exhaustive analysis. Sample about a third of your options before committing, keep what matters close, do one thing at a time, and let good enough be good enough. Read this if you’ve ever been paralyzed by too many choices and suspected there was a smarter way to decide than sheer force of will.