Algorithms to Live By

Algorithms to Live By: The Computer Science of Human Decisions by Brian Christian, Tom Griffiths.

Cover page of book Algorithms to Live By: The Computer Science of Human Decisions by Brian Christian, Tom Griffiths

All models are wrong. But some are useful.“–George Box

This book provides detailed exploration of some models or algorithms as applied to daily problems–ranging from marriage decision to design of parking lots. Simultaneous exploration from two standpoints–computational and psychological–brings two interesting insights.

First, it makes very clear that all theory and practice could be very different. As authors noted “Communication is one of those delightful things that work only in practice; in theory it’s impossible.” It doesn’t stop here and go in details what and why differs. For instance, in chapter on explore-exploit dilemma it turned out that people tend to over-explore. We switch to explore mode faster than algorithm suggests. However, it makes perfect sense if one take into account that model assumes unchanging world. If environment changes–as it constantly does–switching to explore is a good approach.

Second, it factors in computational burden. This could be crucial for explaining and design. The trick is seemingly similar things use different algorithms with different computational burden. For instance, it turned out that people are more likely to be available when you request meeting with specific time (or a number of options), rather than “at convenient time next week.” The former constrained problem is easier to solve. The latter launch resource consuming scheduling algorithm.

Overall the book is an excellent reading. It combines in academic arguments, everyday observations, anecdotes and philosophical pieces.

P.S. I don’t remember exactly how this book came to me. I believe, it is coming from someone reading list.