Page 63 - The Real Work Of Data Science Turning Data Into Information, Better Decisions, And Stronger Organizations by Ron S. Kenett, Thomas C. Redman (z-lib.org)_Neat

Root Out Bias in Decision‐making 51

One preconception that too many data scientists make, particularly those early in their
career, is that they must “fall in line.” No matter what, they must never challenge the boss.
Although there are certainly company cultures that discourage doing so and vindictive bosses,
it is the sort of preconception that you should examine (see “Bringing the Boss Bad News”).
Why do you feel that way? Is there any data to support it? Are there counterexamples?
The second part advises that you reverse your inclinations: what would happen if you
decided to move forward in the opposite direction from that you originally chose? Gather the
data you would need to defend this opposite view and compare it to the data used to support
your original decision. Reevaluate your decision in light of the more complete data set. Your
perspective may still not be complete, but it will be much more balanced.
In parallel, ask yourself, “Am I the right person to decide here? Or should someone else,
who has time to assemble a complete (and hence less‐biased) picture, make this decision?” If
so, then you should assign the decision to that person or team.
Third, before you commit to announcing,
executing, and defending your choice, first Bringing the Boss Bad News
try out your decision on a “friendly” or two. Many people fear telling the boss their con-
A friendly usually refers to someone who is cerns that he or she is about to make a mis-
on your side and wants you to succeed. take, or they fear bringing bad news. After
Here, we are referring to someone who all, the boss may “shoot the messenger.” Our
also wants to protect you and has the experience is that it can be different. We
courage to tell you honestly when your have found that most senior managers are
thinking is incomplete, when you have well aware that people do not like to do this.
missed something important, and when you So they value people who will tell them the
are just plain wrong. If a friendly advises hard truths. So, of course, we are not advising
any of these things, then start anew, com- that you say, “Boss, that’s dumb,” in front of
pletely rethinking your decision and the the entire team. Instead, learn how to raise
data you need to make it. your concerns or bring bad news in a discrete
Few people set out to make a rigged and supportive manner.
decision, but when you are pressured to
make a choice quickly, you may fall victim to a flawed process. By admitting your own precon-
ceptions, and subjecting your thinking to someone who will really challenge you, asking your-
self tough questions, you can expose a rigged decision‐making process. You realize just how
difficult it is to completely eliminate bias. And you will make yourself a better data scientist.

Solid Scientific Footings
To develop a deeper appreciation for a scientific framework for handling bias in decision
making, data scientists should study the groundbreaking work of two psychologists: Amos
Tversky and Daniel Kahneman. Tversky died from leukemia in 1996 at a relatively young age,
while Kahneman was awarded the 2002 Nobel Prize in Economics. They established
behavioral economics, a new domain of great importance to data science. See Lewis (2017)
for a popular description of the work.
A brief summary: we know now that decision‐makers are affected by several mechanisms
that blur their ability to properly interpret data‐driven reports, including:

1. base rate neglect
2. overconfidence