In 1974 a researcher named Paul Slovic studied professional horse bettors. He let them each choose 5 pieces of information to make their predictions. Whatever they wanted from an extensive database. Horse’s age, jockey weight, etc. The bettors could each have a custom list of 5 variables.
The bettors turned out to be nearly 2x as good as chance in predicting winners 17% vs 10% since there were 10 horses running. Furthermore, the betters assigned a 19% confidence to their bets suggesting they were also well-calibrated.
As betting rounds continued, the bettors were allowed more pieces of information culminating in a final round where a bettor was allowed 40 variables.
What happened? Access to more variables had no effect on their accuracy but doubled their confidence. (Source: Adam Robinson interview)
1. Humans struggle to process too many sources of information effectively.
Psychologists and common sense suggest we simply can’t hold too much in our working memory. Since storage is upstream to processing it feels safe to conclude we can’t integrate too many variables in our reasoning.
2. Excessive information can lead to false confidence.
Humans are slaves to narratives, most of which are not rigorously tested. In Moneyball, old school scouts thought having a six-pack was predictive of a player’s success. This might be true if you are choosing from random people and have no other info. But when you are looking at a narrowed pool of baseball players, the sight of impressive abs loses its weight relative to the information file that was used winnow the candidates in the first place. The excess info can not only distract us from our proven model but, and this is diabolical, increase our confidence.
Implications for Reading
When reading to learn, as opposed to entertainment, search for the books that matter. The lessons that generalize. The handful of variables that do most of the explaining.
Probabilistically speaking, the non-fiction you picked up at Hudson News in Terminal 3 is best used to lined a birdcage. The canon that has stood the test of time gives you perspective. You need perspective to judge. Perspective shifts in increments, while the latest study simply gets written over. A draft better off never seen.
Instead of filling your brain with the facts most likely to expire, let’s optimize our reading time.
What to read
Bill Gates has called Charlie Munger, Warren Buffet’s partner at Berkshire Hathaway, the “broadest thinker he’s ever met.” Heavily influenced by Ben Franklin and his almanac, “Poor Charlie’s Almanack is a collection of Charlie Munger’s best advice given over 30 years, in the form of 11 speeches given as commencement addresses and roundtable talks.” Allen Cheng’s notes are outstanding and sectioned by topic. (Link)
While I prefer to be cautious when drawing lessons from outliers such as Munger, his emphasis on multidisciplinary learning bears repeating.
By collecting a broad range of mental models you can more effectively pattern-match phenomena in one discipline with more general phenomena that you may have discovered in a different field. Farnum Street describes 109 handy mental models to add to your repertoire. (Link)
How will they help?
1. Improve your analysis and more importantly predictions
Many phenomena are just instances of a more general one. To analyze the phenomena at hand, we can study how the general one usually plays out. The general one will have a long history and wider sample set to draw from. Consider a hardware store and a candy shop. They seem very different but both are straightforward examples of brick and mortar retail. The underlying business principles are going to be more similar than the difference between Skittles and a chainsaw. In essence, we are collapsing many data points into the ones that mattered. Rather than study every type of brick and mortar we can reference our mental model for how brick and mortar businesses work. Remember the horse bettors. Isolate the information that matters.
2. Avoid misdiagnosing a problem
When you are too narrow, you are the proverbial hammer to which everything looks like a nail.
Example in action
According to The Power of Mathematical Thinking, there was a year in which North Dakota had the lowest cancer rate and South Dakota had the highest. Explain how that can happen.
Where To Start
To get a start or reboot of a personal multidisciplinary education Naval Ravikant offers suggestions:
- Stick to science and the basics
- Generally, read the things that people agree on: “The longer something has been in print, the longer it will remain in print and the higher value it is.” What Nassim Taleb calls the “Lindy Effect”.
- Read lots of microeconomics (NOT macroeconomics)
I’ve offered my own advice in the past. Avoid the news and information with a short half-life. (Link)
How to Read
There are different types of reading which depend on your intent. If you are not reading for entertainment or scanning for a specific piece of info, then you are reading for understanding, retention, and reuse. Mortimer Adler describes analytical and synoptical reading. (Link)
You must read actively which takes effort if you expect to absorb the information for later use. Some techniques for doing this include:
- State what a chapter or book is about with the utmost brevity.
- The Feynman technique — “the best way to learn anything”. (Link)
- Cedric Shin’s Land and Expand strategy for fully conquering a topic. It assumes you are too tired to read difficult topics after a long day at the office. (Link with my highlights)
- The “Barbell”. Nassim Taleb and Marc Andreesen focus on very short term and very long term expiring information. Think social media and classic books.
- Dave Perell plunges into “atemporality”
- For days or weeks at a time, I will escape the present moment and only consume content published in a different decade. For example, if I want to learn about the 1970s, all my media consumption will consist of books, videos, and interviews published in the 1970s. By doing so, I’ll embody the mindset of people in a bygone era and gain new perspectives on the here and now. (Link)