I wasn’t predisposed to love Algorithms To Live By, a new book by Brian Christian and Tom Griffiths that suggests approaching life decisions like a computer scientist.
With the greatest respect to the computer scientists I know, it’s a job that evokes certain cliches not associated with healthy work-life balance, social skills or high tolerance for sunlight. Open the book at random, and you might find that stereotype confirmed.
Did you know that, according to maths, you should marry the first person you meet once you turn 26 who’s better than all previous people you’ve dated? (This assumes you started looking for a spouse at 18 and want to find one by 40.)
Of course, nobody could ever bring themselves to live so mathematically, even computer scientists, and yet, by the end of the book, I was convinced. Not because I endorse the idea of living like some hyper-rational Vulcan, but because computing algorithms could be a surprisingly useful way to embrace the messy compromises of real, non-Vulcan life.
Computer science, Christian and Griffiths point out, is all about coping with limitation. We ask computers to do a million complex things, and at lightning speed. But they have limited processing power, so it’s always a matter of tradeoffs. When is it better to be fast than accurate, or vice versa? When should a computer stop searching for the perfect solution to some puzzle and use a rough-and-ready one instead? Slightly rephrased, these are the central challenges of life.
When do you stop searching for a better partner, flat, group of friends or career path ? You’d like to make the best possible choice, but gathering data comes at a price. Spend your whole life auditioning new spouses, friends or jobs, and you won’t have spent it well.
The best algorithmic solutions vary according to the scenario. One appealing idea, when you’re facing a fork in the road, is to choose the option with the highest “upper confidence bound” – the one that could plausibly perform best in future (even if it’s also the one with a higher chance of being awful). There’s also the useful method known as “constraint relaxation”, a technical way of describing self-helpy questions such as, “What would you do if you weren’t afraid, or if money were no object?”
Examine your predicament, then remove one of the constraints – money, time, family disapproval – and ask what you’d do. The answer may clarify your real-world decision.
But the authors’ most immediately useful concept may be “computational kindness”. When making a plan with a friend, it feels polite to say you’re flexible about when to meet, or that you don’t mind at which restaurant.
But refusing to state your wishes imposes a computational cost on the other person: now he or she must make the choice (while guessing at your preferences). For humans, as for computers, deciding makes demands on limited processing power. Don’t overtax yours, and don’t force your friends to use theirs on your behalf. — Guardian News and Media