Book of the Week: Simple Rules

14 Oct 2015

simple_rules I didn’t really like Simple Rules. One thing that bugged me was when the authors said simple rules outperform complex models, because of overfitting problems. Anybody who builds complex models for a living, should know how to handle overfitting. That’s a bad argument to make. Machine learning algorithms outperform most heuristics now. The book is fairly recent, so some of the things it references feels like it happened yesterday. I tend to think of simple rules are for people who don’t understand what’s happening. In chemistry they teach you rules about how bonds work, but you always encounter exceptions that run contrary to your rules. If you knew quantum physics, you would know why certain things happen contrary to the rules. It’s probably not worth the time to teach all chemistry students quantum mechanics, so you give them a simple rule to follow, which is good enough to get them through med school. Complex Problems Warren Weaver, a director at the Rockefeller Foundation described science progressing through three types of problems: simple, uncertain and complex. For those familiar with material science, you can think of it as the microscopic (simple), macroscopic (uncertain) and mesoscopic (complex). The microscope deals with understanding individual particles. The macroscopic deals with distributions and collective probabilistic behavior. The mesoscopic deals with the middle ground, where theories from the other two don’t work well. Most of my academic research is in the complex region, because the other two are solved problems. Physics can describe a single hydrogen atom well. We can also calculate the properties of a periodic material. What we can’t do well is describe what happens when you have an atom next to an infinite structure. This taxes the limits of current computers. Simple rules are designed to handle the complex problems. Simple Rules Simple rules do 3 things well.

  1. confer flexibility while maintaining consistency
  2. produce better decisions with limited time and information
  3. allow members of community to synchronize

Simple rules have 4 traits.

  1. limited in number
  2. tailored to person or organization using them
  3. apply to well-defined activity or decision
  4. provide clear guidance while allowing one to exercise discretion.

The 6 types of simple rules.

  1. boundary - deciding between two mutually exclusive alternatives
  2. prioritizing - rank options when allocating resources
  3. stopping - when to reverse a decision
  4. how-to - guide for executing tasks
  5. coordination - getting something done with multiple actors
  6. timing - getting things done in situation with temporal factors.

There are 4 ways of forming rules.

  1. Codifying personal experience.
  2. Drawing on the experience of others.
  3. Distilling scientific evidence.
  4. Negotiating an agreement.

The 3 things that happen when rules improve.

  1. shift from superficial and convenient rules to strategic and abstract ones that are more effective over a broader range of activities and decisions.
  2. different types of rules are learned in specific sequential order.
  3. simplification cycling where number of rules grow, then shrink and finally remain constant.

The 3 steps to putting a strategy of simple rules into place.

  1. Figure out what will move the needles.
  2. Choose a bottleneck.
  3. Craft the rules.

Simple rules can be used to make better decisions for organization outcomes and also personal outcomes. The book gives a lot of example of these. The 3 obstacles to simple rules.

  1. effort - takes time and iteration to get right
  2. people who benefit from complexity - lawyers and lobbyist benefit from complexity
  3. “myth of requisite complexity” - thinking complex problems require complicated solutions.

My life is too complex, I should find my own simple rules. Purchase Simple Rules on Amazon.com or check it out from your local library.