In part I, we have seen that error is made of two components: bias (the distance to the truth) and noise (the variability in measurements). In part II, we revisited these concepts in a more formal approach. Now, in part III, we will examine measures to mitigate them.
First, let us establish that cognitive bias leads to both bias and noise. The term “cognitive bias” has become an umbrella term for many psychological mechanisms, which sometimes lead to bias (systematic error) and sometimes lead to noise (random error).
Therefore, debunking cognitive bias is a sure way of reducing both systematic and random error. However, noise requires a special focus: Its stealth nature makes it much more difficult to fight.
Debunking cognitive bias
Availability: one tends to consider as true what one sees a lot. This form of cognitive bias is well-documented and is at the core of most social media criticism. By exposing users to other users that think likewise, the group ends up exposed to more and more content aligned to its beliefs. In turn, the users believe that what they think is true. Although what is true and what is not is another debate, exposing users to what holds predictive power on a given question can help fight this bias. By mixing points of view from different strata and weighting them according to their track record, we can use this availability bias to create smart echo chambers. (You can read more about this in our post “Transforming echo-chambers and availability bias into useful analytical tools”).
Confirmation: one tends to select and interpret only information that goes in the direction of what one already thinks. Fighting confirmation bias is not complex, but it is very difficult. Being right feels so good that your brain will resist being exposed to the remotest possibility of being wrong. The solution, however, is just as follows. Always consider your hypothesis as wrong. Once you formulate it, go out into the world actively looking for information that disproves it. This mind trick is central to the scientific method and can be institutionalized in a team with sparring sessions.
Bandwagon effect: one tends to think what the rest of the group is thinking. Competitions such as forecasting tournaments are great to debunk this type of bias. By creating a sport out of being right, a system will emerge to fight the bandwagon effect in a relatively efficient manner.
Exposing noise
By exposing noise, we usually discover where to spend more analytical time. Here, we will focus on two main noise indicators: Disagreement (when different people have different points of view on a given matter) and inconsistencies (when people disagree with themselves).
Disagreement: when different people have different points of view on something. Assume you need to hire an employee. You want somebody who is organized, so you create a series of questions to check whether the applicant has the desired skill. Two senior managers assist you in the meeting. After the interview, the two senior managers debrief and realize that they strongly disagree on the candidate’s level of organization. One interprets the fact that the candidate works 2 jobs and 3 extracurricular activities as a feat of organization. The other manager sees it as dispersion at best and focuses on the fact that the candidate did not take any notes. Exposing the disagreement allows us to dissect the issue and articulate its definition. An anticipation tournament, in which the positions of different people are plotted on a single chart, allows us to detect where the noise lies at a glance.
Inconsistency: when somebody is in disagreement with their past behavior.
Pretend you are a judge who is almost always more severe than your peers. However, sometimes, you are much more lenient than all of them. Studies show that this could happen due to a gazillion external factors: How much time has passed since your last meal, whether it is sunny or rainy, whether the face of the convict reminds you of someone you like, etc. In such cases, a personal log makes all the difference. Being able to access and work your own data helps you to understand and, if necessary, correct your opinion. Essentially, keeping a journal allows you to balance two risks: not enough consistency on one side, and too much on the other. You do not want to be in a position in which you are unable to change your mind either. But that's something we will explore in another post.
コメント