Your Judgment Is Flawed: It Is Unfair, It Puts People At Risk, And You Don’t Even Realize It.
Noise is variability in judgments. It is when judges give different sentences for same case. It is when doctors give different diagnoses for the same patient. It is skewed performance reviews or disagreements from unstructured interviews.
Noise in judicial sentencing impacts years of human lives. In medicine, it could be a wrong prescription or an unneeded surgery. One story in the book is about a person tagged as a security threat because of a misjudged fingerprint. In these contexts, noise is a big source of unfairness. Institutions can be unaware that they are giving out judgments that are closer to lotteries. The authors warn that people don’t know they have a noise problem, and they don’t know the extent of its impact.
As the authors note, Noise is a component of Error. Error is the combination of both Bias and Noise. Both are big problems. However, while Bias is more noticeable, Noise is hidden. And their overall finding is that wherever there is human judgment, there is noise. and there is more of it than you think.
Sources of Noise
The 3 main sources of Noise are Level Noise, Pattern Noise, and Occasion Noise.
- Level Noise is the variability of average judgments by different individuals. This is when some judges are lenient, and some strict.
- Pattern Noise is the variability of personal responses to different cases. This when a judge is more forgiving for a category of cases, and then harsher for another type of cases.
- Occasion Noise is the variability of judgments for the same case in two different occasions. This is when a judge gives different sentences for the same case presented on any two different occasions, like a Monday vs a Friday.
The biggest source of Noise is “Stable Pattern Noise” or people’s differences that generally doesn’t change over time. Differences such as personality, beliefs, and perspectives.
The Goal: A Less Noisy World
Simple models and rules are better at decision-making because they are noise-free. Machine Learning Algorithms are the best at noise-reduction because they are not affected by different occasions nor by personal preferences.
Critics raise an argument that an over-reliance on rules mean you are tamping down intuition, and people’s individuality. The authors argue that the goal of judgment is accuracy, not creativity. Judgments are not the place to express your individuality. Also, Noise-reduction for the sake of noise-reduction is not advised because you could be excluding a lot of other people, just as you would when an algorithm is built with biased data. The goal is to create systems that are noise-free, and less biased.
Decision Hygiene
To reduce noise, the authors suggest the concept of Decision Hygiene. This prevents unseen forces from introducing randomness in your decisions. Whereas handwashing is a preventive measure for diseases, Decision Hygiene is a preventive measure for noise. It limits your exposure to potential psychological biases.
Examples of Good Decision Hygiene are:
- Create Checklists
- Do not give unneeded information to judges so as not to bias them, even if the information is accurate
- Break down judgments into smaller units and deal with those units independently
- Delay intuition until you’ve looked at all factors
- Get a lot of independent judgments and aggregate those judgments
- Consider problems from reference points rather than as a unique case
Why read Noise?
You learn how to improve your judgments. You learn how to improve accuracy. You learn that by reducing Noise, you reduce unfairness, and suffering and unnecessary costs.
How can you use this information?
Recognize the presence of Noise. Practice Good Decision Hygiene. Create simple rules for recurring judgments.
See more:
Like, Follow, and Get notified in advance through social media sites:
IG – https://www.instagram.com/booknewb/
FB – https://www.facebook.com/booknewb/
TW – https://twitter.com/booknewb
book newb 
Leave a comment