Algorithms to Live By: The Computer Science of Human Decisions in Science and Inquiry

Oct 26, 2017 David Rubenstein rated it it was amazing

Shelves: computers, mathematics, psychology, management, nonfiction

I enjoy thinking about algorithms as they are applied to technical problems. So, when I saw this book, I thought, "This is a book written just for me." And, that assessment was absolutely correct. It is a fascinating book, all about how sophisticated algorithms are applicable to everyday problems.

The book starts out describing the "optimal stopping problem." It is also sometimes called the "secretary hiring problem", and I have seen it applied to dating to find a romantic partner, and this book I enjoy thinking about algorithms as they are applied to technical problems. So, when I saw this book, I thought, "This is a book written just for me." And, that assessment was absolutely correct. It is a fascinating book, all about how sophisticated algorithms are applicable to everyday problems.

The book starts out describing the "optimal stopping problem." It is also sometimes called the "secretary hiring problem", and I have seen it applied to dating to find a romantic partner, and this book points out that it can also be applied to looking for an apartment in a seller's market, or searching for a parking space. It can be summed up by the phrase, "Quit while you're ahead." Basically, you have a finite amount of time to complete a search for the "best" item or person. What do you do? It can be shown, mathematically, that you should conduct a search for 37% of your total available time, without making any decisions. Then, at the very next time you find an item or person that is better than all of the previous ones, you should jump on it! I read about this, beforehand, but I didn't realize that 37% comes from 1/e, where "e" is the natural exponent.

The next topic in the book is the dichotomy between explore and exploit. That is to say, when you are looking for something to do, will you try something new (explore), or repeat something that has pleased you in the past (exploit). The answer has to do with how much more time you have in this situation. Movie sequels are all about exploiting, because movie revenues are decreasing. If you are new to an area, you will explore restaurants, but if you are moving away soon, you will return to restaurants you have previously visited. I enjoyed how this issue is related to the "multi-armed bandit" problem. This is a mathematical problem that has lots of applications; choosing the best slot machine in a casino, for example, or designing a web site that will maximize profits for a company.

Sorting is a big topic in the book. It turns out that top poker players sort themselves out, and don't want to play with players they think are better. Libraries use an inefficient caching system, putting recent acquisitions in front. Instead, they should put the most popular books in front. The Naguchi filing system involves returning all files to the extreme left. This goes against the recommendations of efficiency experts, who recommend sorting by topic. Leaving a pile of papers on top of a desk is an example of the MOST efficient filing system! I enjoyed reading an anecdote about Barack Obama visiting Google, before he became president. At Google, he was treated to a pretend job interview. He was asked, "if you had to design an algorithm to sort a million 32-bit integers, what would you do?" He replied that he would not use a bubble-sort algorithm, and all the people in the audience applauded his correct answer!

I enjoyed reading about human memory from an algorithmic point of view. Memory recall is a problem of organization. It was fascinating to read that a graph of the probability of remembering something as a function of elapsed time exactly mimics the probability that a word appears in a newspaper headline twice in a row!

Knowing more makes it harder to remember things; what we call memory decline is actually just learning. A brain fart is actually just a cache miss!

Related to sorting are various problems of scheduling. Examples from the book include multiple laundry loads to minimize total washing plus drying time, and minimizing the rotting of food from a CSA (Community-Supported Agriculture).

No book on everyday applications of algorithms should bypass Bayesian reasoning, and this is an important subject in this book. Our memories of images of plane crashes and car crashes are roughly the same; as a result, we are surprised that there are orders of magnitude more deaths in car crashes, than in plane crashes.

The dangers of over-fitting are discussed at length. As an example, police and FBI agents sometimes over-train; they sometimes use good gun-training etiquette during a shoot-out, with fatal consequences. I really appreciated the discussion of how over-fitting is avoided in biological evolution. It is dangerous for organisms to evolve to over-fit an ecological niche, because the species might not be able to rapidly adapt to an uncertain future environment.

I also thought that the description of "exponential backoff" was very well treated in the book. This esoteric-sounding algorithm is used in all computer networks, but is also used by many of us in treating flaky friends, and in punishment for probation violations.

The last chapter in the book was about game theory. It is helpful, in reading this book, to have some previous background, an understanding of a "Nash equilibrium". The following quote sums it up nicely:

Love is like organized crime. It changes the structure of the marriage game so that the equilibrium becomes the outcome that works best for everybody.

I highly recommend this book. Regardless of your mathematics background, it will intrigue you and amaze you, to see how math permeates our everyday lives. ...more

126 likes · Like · see review

Sep 13, 2016 Mitchell Friedman rated it it was amazing

Shelves: bc-tu-social, author-c, five-star, nonfiction, favorites, published-2016, five-star-non-fiction, library, psychology, liked-reviews

A re-read. Definitely this was worth revisiting. I found some of the later chapters less impressive than I remembered - mostly because they were more computer science and less algorithms that you could act on. It was fun to try to discuss this with a mixed work group. Certainly the math/computer science/software engineering pushed some of my coworkers a little over the edge. 4.5 of 5.

This is one to buy re-read and to some degree commit to memory. It is a great walk between human life and psychol ...more

4 likes · Like · see review

Feb 13, 2020 Barb rated it really liked it · review of another edition

Shelves: own-it

This was interesting. There were a lot of comparisons between how humans do things and how computers do things. Not enough of this was helping me make decisions, schedule my day, & sort socks!

2 likes · Like · see review

Dec 21, 2016 Clay rated it really liked it

Shelves: non-fiction

Terrific read on how computer science attempts to model and figure out very real human problems. In a sense, I feel like the title is a bit misleading. I feel like it implies that it will teach you how to make "good" decisions, and it does, though I feel like it focuses more on the algorithms at work within social psychology and how humans make decisions independent of optimized algorithms. However- a very fascinating book. ...more

Like · see review