Podcast: Accuracy and precision
Not the same thing
One of the nice things about running the podcast alongside the newsletter is the chance to revisit past concepts and introduce them to newer subscribers. Even if you don’t listen to the podcast, hopefully sharing them here gives you a new sketch to think about each fortnight. I also often use it as an opportunity to revise and tidy older sketches.
This week Rob, Tom and I dove into Accuracy vs Precision, which, surprisingly to me, are not the same thing.
Listen on: Apple · Spotify · YouTube
Or, as an experiment, right here:
Accuracy is true to intention and precision is true to itself I learned from Simon Winchester’s book, Exactly: How precision engineers created the Modern World. Huge improvements in the precision of our production led to all the wonders of modern manufacturing where we expect everything to be near enough identical every time. More in the podcast.
This image is almost exactly what Machine Meaning folks (such as myself) use when they talk about Bias vs. Variance (look it up, the first result will be an image such as yours). Thanks for sharing this different point of view! I've never seen these terms used this way, in Machine Learning accuracy and precision also have precise definitions but they're not quite the same. In short, accuracy is how many of your answers are correct (in a classification setting), precision is how many retrieved documents/links are relevant (in a retrieval/search perspective, often compared to recall: how many of the relevant documents were actually retrieved).
I thought this might be relevant, I'm sure more people will think of this perspective when they see your image! More evidence of the power of simple but effective images, I suppose! (I actually found your project because it was used as an example of effective analytic storytelling during my PhD)
Cheers from the Netherlands
(and yes, last week I exactly came home from some "uitwaaien" when I saw your sketch about it... Right on!)