Posted on: Friday 22nd of February 2013
As Open Data Day is showing the principles of Open Data are gaining ever broader acceptance, along with fast developing practices and practical applications. The potential benefits are huge. Ctrl-Shift is researching a White Paper on what happens when Open Data meets personal data right now.
But while society is making progress on one front, we could be losing ground on others. Two dangers stand out.
Faux openness is the trick of making things harder to find by hiding them ‘in plain sight’. The classic example of faux openness is small print. It’s there for everybody to read. But as small print producers know, the more small print there is, the less likely it is to get read. Being ‘open’ is a very good way of hiding something.
That’s how Google ‘announced’ its decision to abandon its commitment to the principle of natural search and include only paid-for ads in its shopping site, for example – via the small print.
Hiding things in ‘open’ small print brings an added benefit to the person doing the hiding: “We’ve been open about it,” they can say, “so you’ve got no grounds for complaint”.
This is also the problem with the flawed concept of informed consent. Things shouldn’t work like this. When we use a product or service we should ‘just know’ that it is safe, not manipulative, not biased, does what it says on the tin, has the ingredients it claims to have, and so on. We shouldn’t have to read acres of small print to be ‘informed’ about such details, nor should we have to ‘consent’ to them. The apparently empowering process of informed consent disempowers consumers in two ways. First, it creates extra, unnecessary work for them. Second, it lets producers slough off responsibility for the quality of what they produce, to dump it on the shoulder of consumers.
Riding under the banner of ‘transparency’ it actually achieves the exact opposite.
One of the wonderful things about our information age is the wealth creating potential of algorithms – their ability to drive automation into all manner of things, producing results in seconds what previously could have taken years manually (just think of Google search). Algorithms lie at the heart of the next big growth industry of Personalised Information Management Services that deploy information as a tool in the hands of the individual.
There is just one problem with algorithms. Most of them are not open. They’re closed. Proprietary.
There’s a good reason for this. If we can see inside their workings, we can game them. There’s already an entire mini-industry out there devoted solely to understanding and gaming the algorithms that determine what answers Google search engines serve up, for example.
However, if the rules built into the algorithm are secret they can be manipulated to favour Google or anyone else for that matter. And we don’t know – can’t know – that they are being manipulated because their inner workings are secret. If Google thinks it’s OK to abandon natural shopping for its shopping service for example, what does it feel OK to do with the workings of its core search product. What is natural search anyway? Who gets to define what ‘natural’ means or looks like?
This problem – of secret, proprietary algorithms which make decisions which we cannot see or understand, but which nevertheless have important effects – is not confined to Google. It’s everywhere. Some other examples include:
- What price we are offered when looking at a product or service online – something the Office of Fair Trading is currently looking into.
- What information we are presented with when browsing web sites that use traffic histories to serve up content.
- What marketing messages we are presented with online.
The algorithm is a unique product of the information age and at the moment the algorithm economy is evolving in a completely opaque way. What is the point of having Open Data if, increasingly, the way data is actually used to drive crucial decisions about what information we are presented with and what we are offered is hidden behind closed doors?
Should the algorithms used by firms to automate decision-making be open?
Should they, perhaps, be open at least to some form of independent audit?