Hi ,
There's a phenomenon known as "wet" or "soggy bias" whereby some weather forecasts have been shown to consistently overestimate the chance of rain whenever the true probability is very low (e.g.
5%).
The idea is that, in such circumstances, an exaggerated forecast is more useful than an honest one because there is a greater "cost" associated with being unprepared if it does rain than being prepared for rain if it doesn't.
So, by exaggerating the
probabilities, people are more likely to prepare... that is, of course, until they realise the truth. At this point, any faith people might have had in the forecasts is gone for good.
To test how people truly feel about exaggerated probabilities, last week I asked readers whether they would prefer an accurate weather forecast of a 5% chance of rain or a biased estimate of a 20%
chance of rain (given a true probability of rain of 5%).
Around 80% of respondents said they would prefer the accurate estimate.
Here's the thing...
Mark Twain famously once said: "There are three kinds of lies - lies, damned lies and statistics."
And anyone who has been doing data science for long enough knows how easy it can be to get the data to say whatever it is you want.
But to do so completely misses the point.
Businesses use data to inform decision-making because it is less tainted by bias than human opinion. If you were to reintroduce those biases, business leaders might as well go back to making decisions based on their gut.
Honesty might not always seem like the best policy when delivering results, but in the long run, it usually is.
Tell the truth and explain how to make that truth useful.
It's better than becoming the data scientist who cried wolf.
Talk again soon,
Dr Genevieve Hayes.
