Bias and noise in decision making

We make countless decisions every day, from the mundane to the monumental. Yet, our judgment is often swayed by hidden flaws: bias and noise. Both bias and noise can influence how we see the facts or data that are significant to make decision.

Imagine you’re trying to measure your weight. It is important step to gauge your health and decide your diet. Bias is like stepping on a scale that consistently adds 2 kg to your actual weight. Every time you weigh yourself, it’s wrong in the same direction.

Noise is like using different scales that give you varying readings of your weight. Each scale might be accurate on average, but the individual measurements are inconsistent.

What are bias and noise

• Bias is a systematic error in judgment that consistently pushes our decisions in a certain direction.
• Noise is random variability in judgments. It leads to inconsistent decisions even when the underlying situation is the same.

Some common examples in decision making: Confirmation bias is the tendency to favor information that confirms our existing beliefs. Anchoring bias is overemphasizing the first piece of information we receive. Meanwhile, noise in a performance review, different managers might rate the same employee very differently even though they were using the same performance indicators.

As Daniel Kahneman, author of “Noise: A Flaw in Human Judgment,” explains, “noise is random variability in judgments, while bias is a systematic error.” Noise is invisible since it is inherently statistical. We tend to attribute bad judgments to identifiable causes like biases, while noise is a statistical phenomenon that requires looking at a collection of judgments to detect.

Bias over noise

Kahneman argues that we often overemphasize bias while underestimating the impact of noise. Bias has an “explanatory charisma” that noise lacks. They can be identified and dismantled through logics audits.

Noise does not have catchy names, e.g., anchoring bias and groupthink. Even bigger charisma, are ’logical fallacies’ since people love to throw their latin names in an argument to sound smart: ad hominem, non sequitur, or post hoc ergo propter hoc (I am guilty as charge).

The Good News

Fortunately, most bias mitigation interventions are actually noise mitigation. To go back to the scale analogy, fixing the scales can address both bias and noise at the same time. If we calibrate all scales to be accurate, we remove the bias. By using the same accurate scale, we reduce noise.

In future post, I will discuss my experiences in group meetings and how building a consensus on decision criteria using Multi-Criteria Decision Analysis (MCDA) can be used in mitigating bias and noise.