Algorithms to Live By

Brian Christian and Tom Griffiths, Algorithms to Live By: The Computer Science of Human Decisions (Henry Holt, 368 pages, ISBN 978-1-62779-036-9). Wikipedia dates it 19 April 2016; the routing note that sent it here said 2015, and the page follows the stated date.

Read through descriptions of the book, not the book. The Wikipedia entry is a stub and the book's own site lists no chapters, so everything below about content comes from the one-line synopsis, the site's blurb and the curator's routing summary. None of it was checked against the text.

The claim

Computers and people face the same limit, finite time and finite memory. The solutions computer scientists built for that limit can therefore be borrowed for everyday choices: what to do or skip, how much disorder to tolerate, when to try something new and when to stay with the familiar. The site names the 37% rule (optimal stopping, the "when to stop looking" answer to choosing a spouse or a parking spot), game theory, memory and caching, inbox and closet organisation, and forecasting. The routing summary adds explore/exploit, sorting, scheduling, Bayes's rule, overfitting, randomness and networking.

Why it is here

It is the corpus's first source about normative decision-making. heuristics-and-biases and thinking-fast-and-slow describe where people depart from a standard; this book proposes the standard and treats heuristics as good solutions to hard problems rather than errors. The 37% rule is a mathematical result under stated assumptions, not an empirical finding about people. No study of human subjects is cited in what was read, so the evidence caveat for this spoke applies in full: the algorithms are sound as mathematics, and whether people gain by following them is not shown here.

Reception

Kirkus called it "an entertaining, intelligently presented book for the numerate and computer literate"; Library Journal praised the audio edition; the Berkeley Science Review noted a "sweet optimism regarding human behavior." The site lists bestseller and best-of-year placements. No criticism appears in the sources read, which says more about the sources than the book.

Authors

brian-christian (writer) and tom-griffiths (cognitive scientist). Their affiliations differ by source: the book's site says Griffiths is at Princeton, the routing summary says Berkeley.

Cross-spoke context

Closest neighbours, not ingested there: ../optimization-algorithms-wiki (exploration-vs-exploitation), ../machine-learning-wiki (bayesian-inference) and ../mathematics-wiki (probability and Markov processes).