As the legal chatbot DoNoPay shows, automation may only affect the repetitive parts of white-collar work. The middle classes can breathe again

We are currently going through one of those periodic phases of automation anxiety when we become convinced that the robots are coming for our jobs. These fears are routinely pooh-poohed by historians and economists. The historians point out that machines have been taking away jobs since the days of Elizabeth I who refused to grant William Lee a patent on his stocking frame on the grounds that it would take work away from those who knitted by hand. And while the economists concede that machines do indeed destroy some jobs, they point out that the increased productivity that they enable has generally created more new jobs (and industries) than theydisplaced.

Faced with this professional scepticism, tech evangelists and doom-mongers fall back on the same generic responses: that historical scepticism is based on the complacent assumption that the past is a reliable guide to the future; and that this time is different. And whereas in the past it was lower-skilled work that was displaced, the jobs that will be lost in the coming wave of smart machines are ones that we traditionally regard as white-collar or middle-class. And that would be a very big deal, because if theres no middle class the prospects for the survival of democracy are poor.

Whats striking about this fruitless, ongoing debate is how few participants seem to be interested in the work that people actually do. Most jobs are in fact bundles of different but related tasks. Or, as David Autor of MIT, one of the worlds experts on this subject, puts it: Most work processes draw upon a multifaceted set of inputs: labour and capital; brains and brawn; creativity and rote repetition; technical mastery and intuitive judgment; perspiration and inspiration; adherence to rules and judicious application of discretion.

Typically, Autor argues, these inputs each play essential roles by which he means that improvements in one do not necessarily eliminate the need for the others. And if so, productivity improvements in one set of tasks brought about by automation often increase the economic value of the remaining tasks. This is why, when we consider the possible impact of automation, we should be thinking not of work but of tasks. Having some tasks done by machine might make us more productive in others and keep us in employment.

What brings this to mind is an intriguing website DoNotPay created by a young British student at Stanford University, Joshua Browder. Think of it as a legal chatbot an automated service that provides free legal advice on a number of routine issues. It started out by making it easy to write a letter contesting a parking ticket: you are asked a number of questions (number of the ticket, etc) after which it drafts a letter in the appropriate legal jargon. With parking tickets it claims to have a 55% success rate, so given that its free it looks like a reasonable bet, if you think you might have a case.

Since its launch, Browder has significantly expanded the cognitive and jurisdictional reach of his bot. It now claims to cover upwards of 1,000 different legal issues (from tackling disputes with a landlord to what to do if your credit card is stolen, how to deal with unwanted cold calls, contest insurance claims, extend maternity leave or deal with harassment at work) and suggests remedies that are applicable in all 50 US states as well as in the UK.

Browder calls his chatbot a robot lawyer, but thats not quite right. What it does is to automate some of the mundane, routine things that professional lawyers do writing a cut-and-paste cease-and-desist letter, for example but free of charge, rather than at a price that deters most people and therefore increases inequality. For me, its just drafted an impressive notice under the Data Protection Act 1998 not to use my personal information for direct marketing. Its not rocket science, but as a non-lawyer I might have got the legal terminology in the body of the letter wrong, and I certainly would not have known how to tell the offender that, if he does not comply, I can apply to the court for an order against you under section 11 of the Data Protection Act.

DoNotPay provides a terrific illustration of how technology can be used for socially useful and democratic purposes. More important, though, it also suggests a better way of thinking about robotics and work by making distinctions between tasks that can and should be automated, and those for which human experience, sensitivity and creativity are necessary. Much of what lawyers do is doubtless money for old rope in which case we should not be paying through the nose for those services. We still need lawyers for many other things, for which there is no routine solution and which do require original thinking. So they may wind up poorer; but theyll still have jobs, and perhaps be less bored. And well all be better off.

Read more:

Come friendly robots and take our dullest jobs | John Naughton
Tagged on: