Hello World by Hannay Fry

I recently read this book, Hello World by Hannah Fry. She's a mathematician and this is a book about algorithms and how they interact with humans. And it's 1000% better than that sounds!

The main point being made is that because algorithms are generally better than humans at the very specific task they are designed to do (like recognise cancer cells or faces, or predicting reoffending, or playing chess) people end up thinking they're somehow infallible and give their decisions a weight they shouldn't have. Then the algorithm will inevitably make a mistake, sometimes with dramatic consequences, and people go all all or nothing and want to remove the algorithm from the process all together. But what's best, Fry argues, is if we treat the algorithm as what it is, a calculation we've designed, make the level of uncertainty with which it arrives at decisions much more explicit (for example, for a facial recognition in a crime scene, show the number 2 and 3 matches too) and let the final decision be taken by a human. The best result is when humans and algorithms work together because they're both good at different things.

I really loved this book, it's not a topic I know very much about, and I found it very accessible without being patronising. Maybe that's actually why I liked it so much. It does have all the tropes of non fiction books that I generally find pretty annoying (like picking a single person, describing them by first name and making them emblematic), but for some reason I didn't with this! It helps that she's British so the language and frame of reference is mine - a lot of non fiction books I've read recently have been American and to continually see everyone described as Joe Bloggs, PhD does get pretty grating after a while.

So yeah, I definitely recommend!