Big data

Is there an Uncanny Valley for data?

Louise Marston - 16.05.2012

As I go about scoping some new research on big data, open data and some of the opportunities and challenges for innovation, I've been wondering if there's an equivalent to the Uncanny Valley of robotics that governs how comfortable we feel with the data that we share.

As part of this programme of work, I'm interested in the barriers that stop people using, sharing and innovating with the data they already have and can access. One of the major things that holds people back are concerns with privacy, data protection, and companies concerns about overreaching with data and alienating their customers.

In robotics, there is an idea of the 'Uncanny Valley', which was created by Japanese roboticist, Masahiro Mori. I learned about this idea from an Economist article, but you can also read this interview in Wired with Masahiro Mori.

By Smurrayinchester [CC-BY-SA-3.0 (http://creativecommons.org/licenses/by-sa/3.0/)], via Wikimedia Commons

The idea is that we identify, and feel comfortable with, quite crude representations of humans, like cartoons and stuffed animals. As robots become more human-like (say, from the Roomba to Asimo), we get more comfortable, until the point where it seems 'uncanny' and we treat them as weird and alien.

A Bunraku puppet, an uncanny example from Mori's original chart [from Wikimedia].

I've been wondering if there's a similar effect in play with data and privacy. We like to receive customised discounts on the things we buy, we like online shopping that shows us only the things that will fit, or groceries we've bought before. But there is a certain point where the prediction of our wants and needs becomes too good, and now we feel like we're being followed. This tweet illustrates this idea very well:

other day I admired a coat online didn't buy Now everywhere I go that coat is there on the page Now wont buy the coat on principle

— Gloria Nicol (@thelaundry) April 26, 2012

More spooky is the account in the New York times of Target, the US retailer, using shopping data to identify which of its customers are pregnant, and offering them coupons for baby equipment. If the coupons were all for baby things, there were complaints (not least from a father whose teenage daughter had received them). If they were scattered through a range of randomly selected other offers, then they were accepted.

Although very little we do these days is private, most of the time, we don't like to think about it, and would rather live with the illusion of anonymity but the benefits of personalisation. Supermarkets use vast data and analysis to arrange their stores. Many companies use game theory to analyse decision-making processes, and ensure as many sales as possible. Where does the line lie between good customer service and manipulation?

The question about where the boundary should fall between these positions will define our interactions with data in the next few years. That may mean companies being more transparent about how and why they want to use this data, and consumers getting used to the reality that personalisation and convenience is often achieved by sharing data.