Just Finished Reading: Hello World – How to be Human in the Age of the Machine by Hannah Fry (FP: 2018) [240pp] 

Algorithms are everywhere these days. They’re in ‘self-drive’ cars, they’re in our phones, they offer us ideas on Amazon, Netflix, Facebook and Spotify, they predict the chance of a criminal reoffending, they can even (so it's said) predict crime itself. The problem is, is that they’re often not as good as their developers say they are and its difficult for outsiders to know exactly why they sometimes fail (and sometimes spectacularly fail) because their code is business confidential. So how much can we trust them? 

This is first and foremost not an anti-algorithm book. The author (and I to an extent) agree that well written and well understood algorithms are useful in a whole host of ways – and not just for alerting you to music artists or books that you might not have come across otherwise. When used properly algorithms are very good at cutting through mountains of information and using that data to predict things like cancerous cells in otherwise healthy tissue allowing early intervention and much better health outcomes. When used badly or thoughtlessly, or just accepted at face value without any critical appreciation, they can cause a great deal of harm and even end up killing people as in systems increasingly flying our airliners or deciding on whether or not someone deserves bail. But even when we are being critical that might still not be enough. When an algorithm decides you shouldn’t be allowed to fly or that you should be denied benefits or access to scarce medical or educational resources it's difficult to know why these decisions were made if the algorithm or its developers cannot be questioned – even in court. Likewise, all too often the decisions being made are not even questioned because it's all based on the data, right? Numbers don’t lie, right? Time and again it has been shown that decisions presented on a computer screen are believed over and above the people protesting that the information is incorrect. A typo embedded in someone's data file can take years to correct and may need to be corrected time and again if that erroneous fact is repeated and copied across organisations. Used by unthinking and apparently irreproachable algorithms this can cause no end of problems. 

Overall, this was an interesting look at where our technology is (apparently inexorably) taking us. But what the author rightly cautions us against is both believing the hype and accepting the results thrown up by the sometimes sloppy use and all too easy acceptance of algorithmic number crunching. Recommended if you’ve ever wondered why Amazon sometimes offers you something you wouldn’t buy in a million years or if you want any more sleepless nights wondering about future crime prevention programmes.