Thinking in Bets by Annie Duke
Before we can get anywhere with Data Science, we must embrace "Thinking in Bets." We must conclude that:
- Not everything is predictable – Everything is probabilistic,
- Not everything is entirely chaotic – It is possible to evaluate and manage uncertainty, and
- There are advantages to managing uncertainty.
Unless we believe that we have some limited control over uncertainty, we will fall victim to one of two traps. We will either think we can completely control events or conclude that we have no control over events. Both positions are demonstrably false.
I highly recommend readers get the book Thinking in Bets by Annie Duke. Duke was a professional gambler. She learned through that experience that there are "smart bets" and "dumb bets.” She now applies those lessons to business and general decision-making. She is a business consultant and co-founder of The Alliance for Decision Education (see: About the Alliance for Decision Education | Alliance for Decision Education ).
To use statistical inferences effectively, we must make good decisions in the face of insufficient information. We must set our decision points based on analyzing each decision's costs, benefits, and probable outcomes. This analysis becomes especially tricky when we cannot reduce the costs, benefits, and results to a standard measure such as money. Nevertheless, we should perform this analysis whenever practicable. Although a statistical analysis may not settle an argument, it will focus our attention on the "subjective" issues without conflating the discussion with irrelevant "facts." For example, if we could reduce a debate about widening a "dangerous" highway to something like, "Are you willing to spend $200 million for a 90% chance of reducing highway deaths by 0.5%," we can draw attention to affordability and potential compromise. We also plainly state the consequences of doing nothing and the probability that the proposed decision will not have the desired effect.
If we become committed to “thinking in bets,” we will naturally desire better objective data and employ more powerful decision-making tools. In addition, we will require uncertainty measures to accompany all data to “propagate” that error into predictions we make using those data. Thus, we will naturally ask relevant questions of Data Scientists.
Stites & Associates, LLC (SALLC) is a technology development group founded by Ron Stites in 1996. The scientists and engineers at SALLC strive to help clients succeed at technology development and deployment using Evidence-Based Decision-Making. Feel free to contact Ron Stites via email at [email protected].
Comments