Algorithms To Live By

Some quick thoughts and a review of 'Algorithms To Live By'

‘Algorithms To Live By’, written by Brian Christian and Tom Griffiths, describes a series of algorithms, or more accurately, a series of concepts from fields such as statistics, neuroscience and psychology in order to improve one’s decision-making. The decisions can be as mundane as figuring out the optimal parking space in a busy carpark on one end, and as profound as finding a partner on the other.
These examples are both from the book’s first chapter, which is based on the concept of optimal stopping. What is the optimal sample size of data or examples to have before settling on a particular option? What is the likelihood on finding the optimal option in any particular iteration of searching? It turns out there is a heuristic for tackling these sort of questions, which is known as the 37% rule. The origin of the 37% comes from the ‘Secretary’s Dilemma’, which is about one would determine who is the best possible candidate for a secretary role. Essentially, the optimal point at which to stop interviewing candidates, if there were 100 candidates, for instance, would be the next best applicant after the 37^th applicant. This is known as the ‘look-then-leap’ rule, which gives a clear heuristic in which to stop. It just so happens that this point occurs at around 37% of the way through a search. The size of the pool only marginally affects this rule – whether the size is five applicants or a million, stopping at as close to 37% will yield approximately the same result.

Another chapter which resonated for me was the chapter on scheduling, with the appropriate subtitle, ‘First Things First’. In this chapter, examples are used from organizational science and scientific management, from Taylor’s Management Science, Toyota Lean and Just-In-Time. Crucially, a number of different perspectives are offered, in order to avoid a reductionist, one-size-fits-all approach to scheduling and time management. Not all situations are the same, and using a formulaic response will not result in optimal outcomes, according to Christian and Griffiths. What they advocate for instead, is a modified form of shortest processing time, a heuristic which emphasizes doing the quickest task first as a way of getting momentum. Their modification includes ‘weighting’ each task, in terms of importance, such as how much money it’ll earn a firm to do it, or how much it will help other urgent or strategic goals. Combine the weighting and the length of time it’ll do a task, then work through the most valuable tasks.

To an extent, I was already using many of these heuristics in daily life to some extent, as do many people in some form. What makes the book interesting, however, is that it makes these heuristics explicit and explains how it is they work. Having this all laid out as Christian and Griffiths do in the book makes it much easier to understand the thinking behind it, and to apply these rules more rigorously. It also allows for them to be applied not only in professional situations, as one might expect, but also in everyday life in varying degrees. Crucially, it works with nuance and compromise that one will encounter in everyday life. I tend to be weary of self-help or self-improvement style books, but the angle taken in Algorithms to Live By separates it from most in the genre, and packs a lot of information into a reasonable length. Regardless of if you’re familiar with computer science and statistics or not, this is an eminently readable book, whilst not oversimplifying things. Overall, this is a book I was pleasantly surprised by, and one I would recommend to anyone interested.