An algorithm is just a series of instructions – a set of rules to solve a problem. You can perform one in your head or with a pencil. But if there is one thing computers are brilliant at, it’s following rules, and the most important algorithms today are all performed by computers, and written in programming language that computers can understand.
Type anything into Google and powerful algorithms spring into action to rank each page for you: does it contain the right keywords? How old is it? What else links to it? For predictive policing, they take information on what crime has been committed, where and when, and run it through a sophisticated predictive model and then spit out areas where they think crimes are likely to be committed. The prediction itself is hard to explain, but some have been known to be based on algorithms used to predict the aftershocks of earthquakes.
In practice, there might be thousands of such instructions, constantly changing as algorithms learn. The actual machinations of algorithms themselves are closely-held corporate secrets. Look underneath all the code, all the technology and sophistication, and each algorithm, deep down, is just an idea of how to solve a problem. But nothing in an algorithm is above argument or challenge.
Carl Miller is the research director at think tank Demos in the UK.