Author: Daniel Kahneman
We have all been educated on the inherent flaws and inaccuracies of human judgement. However, the world we live in today, still makes use of this judgement to settle many important issues like judges handing out sentences, insurance adjusters deciding claims, teachers grading students, etc.
The authors break down the errors in these decisions into two components: bias and noise. There's plenty of books written on bias and this is not one of them. The authors spend a good amount of time defining noise and breaking it down into various components like
1. level noise
2. stable pattern noise
3. occasion noise
and then delve into each of these sub species. **Level noise** is a result of variability in the bar (level) that each judge inherently judges against. **Stable pattern noise**, on the other hand, is variability within the decisions of a single judge that arise as a result of some pattern that biases the judge away from his normal level. And **occasion noise** is introduced by the environment, mood and other factors that are specific to the instance in time when the judgement is rendered. The authors are very thorough in their treatment of the different facets of noise and provide many suggestions to reduce them. I found the suggestions to improve hiring decisions and the protocol for conducting strategic discussions to be the most useful.
While the book is well written and highlights an area of judgement that hasn't received much attention, I felt the separation between bias and noise was somewhat arbitrary. The way I understood it is as follows.
- Human judgements are prone to errors. These errors are the result of several components ranging from our inherent bias, to our mood, to other environmental factors and so on.
- If we are able to conduct multiple independent judgements of the same underlying data, we are likely to get different answers.
- Even without knowing what the true answer is, the fact that there is variance highlights that most, if not all of these answers have an error component. The deviation of the errors from the average, is defined as the noise.
- Note that the average value is not the true value. The average value is really the "true value + bias".
A more mathematical and succinct definition is “The fixed offset (mean) of these answers from the true value is called bias, while the random component is called noise.”
While the above definition is technically sound, I don't know if the underlying processes that cause these errors can be so neatly separated. Bias itself can be a random process and if we allow for that, then the definition of noise is suspect. Additionally, a strategy to reduce bias, may result in reducing noise as well. Having well defined rules might be a good way to reduce noise, but the authors point out that it could have the undue side-effect of introducing bias.
Here are some thought provoking quotes from the book
- In general, we jump to conclusions, then stick to them.
- There is so much noise in judgment that a noise-free model of a judge achieves more accurate predictions than the actual judge does.
- Judgment is like a free-throw: however hard we try to repeat it precisely, it is never exactly identical
- Oddly, reducing bias and noise by the same amount has the same effect on accuracy.
- Humans can at most keep seven levels, plus or minus two, for quick evaluation.
- The institutions of justice should acknowledge the limitations of the people who administer it.
- Replacing absolute judgment with relative ones, when feasible, is likely to reduce noise.
- Behaviors are a function of personalities and of situations.
Overall, it's a great book that definitely got me introduced to the concept of noise in our decisions.
