AS A STATISTICIAN, I think our obsession with objective truth and individualism is making us less effective at fighting the pandemic. There's been a lot of debate recently about whether covid-19 is airborne. It's the wrong question! Let me explain...
Imagine we both have a jar of jellybeans. Yours are all red. My jar has a mixture. You like red ones. So logically you prefer your jar of jellybeans over mine. Does this mean that you would prefer a red one from my jar vs a red one from yours? Of course not. They're the same.
How about before we observe the color of the jellybeans. Would you prefer a randomly chosen jellybean from your jar over mine? The answer is clearly yes. You're more likely to get a red from yours. So you see, the act of observing the jellybeans changes what choice is optimal.
In statistics, observations change our model of reality and this can fundamentally change what our optimal decisions should be. In statistics, it's logically consistent to recommend one thing first, but then perhaps recommend the complete opposite with more information.
What's the lesson here? Data is hard. It requires mental flexibility. This phenomena of switching your decision as more information presents itself can happen multiple times with each new bit of information. It can even depend on the order in which you learn the information.
The rapid shifts in public policy recommendations can be frustrating for the general public but it comes with the territory. This is what data-driven decision-making looks and feels like. It involves a lot of changing your mind.
In data analysis, you need to be able to accept there is often no single objective truth. There is only personal truth based on what you know so far and the requirements of your decision making task.
The objectivity of data analysis comes from our decision-making principles, not from the particular decisions themselves.
"Is covid-19 airborne?" is the wrong question. A better question is what is my probability of contracting the disease and how does it vary with the level of social distancing, my protective gear and the prevalence of the disease in my local environment.
Estimates of these probabilities are not fixed and unchanging! They will vary based on the available information about you and the people around!
Is this hopeless? No because there is a single answer to how much masks or any intervention will affect the group collectively. This is the logic behind *public* health. This is why a collective view of health is often more appropriate than a individualist view of health.
This why we need to think about the effect of policies in terms of what will happen collectively when we all follow a rule instead of what will happen to me if I follow this rule because there's no one answer to the 2nd question. There are 7 billion. One for each human being.
If our societies are going to make the transition to data-driven decision-making, we're going to need to get better at thinking in probabilities. We need to stop pursuing simplistic binary answers. Our lives may literally depend on it. Thanks for coming to my TED talk.
