The seductive, dangerous illusions of Big Data

Big Data is a powerful tool. But if we lose sight of the human beings that underlie it, we run the risk of making inhuman decisions.

Like many, I was pretty firmly in the mindset that big data was an unqualified good thing. OK, sure, I had the odd qualm about privacy, but on the whole the idea that more information means better decisions means a more productive world appealed to me. NEXT11’s Data Love made a pretty convincing case for this…

Then this crossed my radar this morning, as I was going through my RSS feeds:

David Brooks, in his Times column today, looks at the rise of “data-ism” — the rapidly spreading belief “that everything that can be measured should be measured; that data is a transparent and reliable lens that allows us to filter out emotionalism and ideology; that data will help us do remarkable things.” Brooks is wary of the worship of Big Data. He worries, wisely, about our tendency ”to get carried away in our desire to reduce everything to the quantifiable.”

On the drive home this evening, the radio was full of the news of a disaster in the British National Health System. The standards of care had broken down catastrophically, with harrowing images of patients being left alone in soiled sheets, or being forced to drink the water out of flower vases, because none of the care staff would bring the patient a drink. What should be a prime example of the benefits of civilisation had become something out of a prison camp or a horror movie. People died, because people employed to care about them failed in their job.

And what was to blame? Allison Pearson in The Telegraph:

When a four-month-old baby was discovered in hospital recently with a dummy taped to his face, it came as no surprise to learn the name of the institution concerned. Between January 2005 and March 2009, the Mid Staffordshire hospital was a terrible place, where patients were left lying for hours in their urine and faeces. Those who tried to get out of bed and help themselves often fell, and their injuries went unrecorded. You see, old people drinking out of flower vases because nurse won’t bring water are of no concern. Not if you want to hit your targets and achieve the holy grail of Foundation Trust, which Mid Staffordshire most definitely did.


But maybe that’s just the view of that right-leaning newspaper. How about the left-ish BBC?

The “appalling” levels of care that led to needless deaths have already been well documented by a 2009 report by the Healthcare Commission and an independent inquiry in 2010, which was also chaired by Mr Francis.

They both criticised the cost-cutting and target-chasing culture that had developed at the Mid Staffordshire Trust, which ran the hospital.

Suddenly, I was taking the concerns raised in the earlier article much more seriously. Measurement and targeting were introduced into the UK’s hospital system for a very good reason – to improve performance. The law of unintended consequences is a terrible and powerful thing, though, and in this case the net effect was to focus people’s minds on the data, not the people the data represented. It, to use Brooks’s expression, filtered “out emotionalism” – in this case, the very emotion, the very empathy that should make hospitals caring places.

They started serving and paying attention to the data, and not the people. 

When we forget that our data sets illuminate the lives of real, living people; when we hide behind targets and jargon like “social graph”, we run the risk of forgetting that humans generate that data, humans use that data, and humans make decisions about other humans based on that data. Data is powerful, but its most powerful in context. And the human context is the one that matters.

Here be dragons.

Photo by Kenny Louie on Flickr, and used under a Creative Commons licence