The dangers of Big Data and the biases of interpretation

Health services in South West London are in the process of being reconfigured. Which means the A&E department and maternity ward at my local hospital could soon close.

Residents have been told not to worry as it will only take 13 minutes to travel to the reallocated wards in Tooting.

How was the 13 minutes calculated? We aren’t told. Cut the data by traffic-dodging, siren-blazing ambulances and perhaps it is an accurate reflection. But does it factor in peak-time traffic? Does it accurately reflect the travel-time for visiting relatives or women in labour who have no sirens or means of dodging traffic?

The data has been cut to show what it needs to show and interpreted with underlying biases.

Which is why we should be welcoming Big Data with open arms and a healthy dose of critical thinking.

The risk of human bias in both the cutting and analysing of Big Data remains just as real – as does the danger of attributing cause to effect.

Take, for example, the Twitter and Foursquare data captured around Hurricane Sandy. In a recent article for Harvard Business Review, Kate Crawford, dissects a study into the real-world implications of social data. Crawford reveals that the data captured during Hurricane Sandy indicated a grocery shopping peak the night before the storm – no doubt people stocking up – with a peak in nightlife the night after – perhaps people letting off tension.

The data also indicated that Manhattan was at the centre of the disaster – with significantly more tweets about the storm coming from the city. What this data doesn’t factor in is the higher level of smartphone ownership or Twitter uptake in Manhattan, comparative with other, worse hit locations.  Or the impact that power blackouts, battery outages and limited phone access had on the ability for those worst hit to convey their experience of the storm on social media.

In short, data can and does often paint part of a picture. It’s up to the interpreter whether he chooses to accept that portrait as conclusive, or whether he continues searching for the bigger picture.

 

Latest Posts

B2B leads go cold when interest is captured before the buying group is ready to move. A form fill shows that someone acted, but it does not mean the decision is ready. Social keeps the commercial conversation alive by carrying proof, building trust and showing what buyers are researching before sales can see it.
Read More
Creative content on social media tends to fall into two camps: the stuff you actually remember, and the stuff you clicked on once… then instantly forgot. We’ve all experienced the second one. The extra dramatic hooks, the “OMG, this will TOTALLY change your life” claims, the slightly over-the-top thumbnails. It’s…
Read More
FMCG brands don’t need more hacks. They need to understand the behaviour behind the feed. This is my particular bugbear right now. The algorithm is only useful when you understand the people behind the signals.  We get endless tips about timing, hooks, formats, posting frequency and “what the algorithm wants”,…
Read More