Scientists have this nasty habit of being honest about degrees of confidence, which leaves politicians enough room for mental gymnastics, and allows people with direct conflicts of interest to sound totally confident in their denials.

When confronted with two people, one saying “we could see global temperatures rise anywhere from half a degree C to 2.5 degrees C, and this could lead to sea levels rising 30 to 60 cm, and may increase the frequency and strengths of hurricanes, el ninos, and la ninas” and the other saying “I promise you, none of that will happen, they’re just being alarmists. Also, those numbers seem awfully small to be concerned about anyway, don’t they?”, well…

I see this at work all of the time. I’ve had to change how I report stuff because I used to have conversations like

Management: “Hey, Kichae, we released an A/B test for a new update. Did it increase playtime?”

Me: “The model shows playtime increases of between 20s to 3 minutes per week.”

Management: “Guess not!”

Now I strip the uncertainty ranges, and I focus on reporting for individual cohorts, so people can see that some players went from playing 3 minutes to 4 minutes, and some players went from playing 3 hours to 4 hours. It’s resulted in a lot more work for me, but at the same time, people are less likely to write off meaningful results.