The Guardian view on automating poverty: OK computers? – The Guardian

Across the world, governments are investing in machines that they hope will run their social security systems and other services more cheaply and effectively than humans. The Guardians Automating Poverty series includes reports from the US, Australia and India as well as the UK. The roles played by technology in these countries are all different. But taken together, the articles reveal how automation, machine learning and artificial intelligence are extending their reach into peoples lives through the delivery of public services.

As with all automation processes, speed and efficiency provide the rationale. But our reporting on systems such as those used in the US to collect government debts, and in the UK to administer social security payments, gives cause for concern on several grounds. These include practical questions such as whether the new systems work, and particularly whether they are equipped to rectify errors and false results.

Given the suffering that even a single missed payment can cause to a vulnerable claimant, any glitches in such systems must be taken seriously. In the UK, there are numerous instances of universal credit (UC) payment problems linked to automation. One man ended up homeless after a computer stopped his payments on the basis of flawed data, a decision that staff appeared unable to override. In India, Motka Manjhi died after his biometric thumbprint key went unrecognised, leaving him unable to access the government food rations that he was entitled to. His family blame his death on starvation.

But even were such technical hitches to be ironed out, or sufficient staff retained by government agencies so that flawed machine decisions could be routinely overriden, serious questions remain. One is whether, and in what circumstances, it is right for a government to replace employees with computers in its dealings with dependent citizens. Given that some of those who rely on the state for support because of their age or disability, or some other reason, are often to some degree marginalised or excluded, arguably what they most need from the government is not just money, but also human resources in the form of staff who can help them.

This was the view taken by UC claimant Danny Brice, whose learning difficulties and dyslexia have made applying for benefits online a torment. Talking is the way forward, not a bloody computer, he told the Guardian. With 5 million adults who have either never used the internet or last used it more than three months ago, and millions more who are functionally illiterate, it is clear that even in a country with universal access to free secondary education there are significant barriers to any online service. This is why Labours recent promise to reform UC so that it is no longer digital by default was warmly welcomed by poverty campaigners. It is also one reason why the current governments continued push to digitise, including the Department for Work and Pensions secretive intelligent automation garage project on checking benefit claims, should be scrutinised far more intensively.

Technology is transformative, but its effects are not always beneficial. The predictive algorithms pioneered by US social media and advertising companies continue to be put to destructive as well as creative uses. Much stronger regulation is needed, for example with regard to data collected from children. Machine learning, like computers more broadly, may serve socially useful purposes, including in the design and delivery of public services. But such applications present moral and philosophical challenges as well as technical ones, among them questions of consent, privacy and the replication or even intensification of human biases.

So far, this series suggests, the uses made by governments of the latest automation technologies raise more questions than they answer. The public in each of these countries needs to know much more.

Go here to read the rest:

The Guardian view on automating poverty: OK computers? - The Guardian