Big Data: Is Reputation Really Portable?

At the recent London New Finance session on Big Data in Finance, Mark Hookey of Demyst.data suggested that a more accurate profile of a person is obtained by observing the breadth of the person's behaviour, rather than the depth of their history in any one area. The challenge is knowing which types of data from each area of the person's behaviour are representative (and having permission to use that data). He conceded that the profile is probabilistic rather than predictive.

Rachel Botsman has also talked about the concept of 'reputation capital', which is a product of all who have trusted you, when and why. She says it's only a matter of time before we are able to aggregate, monitor and use our ratings on the many sites on which we interact, so that we extract more value from the total of our "reputation capital". Rachel suggests this capital will be more powerful than our credit score. Rachel also suggests we'll be able to intentionally 'shape' our reputation, and so build-up our reputation capital (or reduce it). Two challenges she suggests are:

• knowing which data should be included in the data set that comprises your total reputation -  the same challenge facing Demyst.data and others Rachel mentions; and

• how to enable 'digital ghosts' to leverage their reputation capital (subject to privacy and data protection), since they don't interact online and therefore do not personally generate their own reputational data. 

But even if you do manage to identify the limited set of data that best represents a person's behaviour in a given context: 

• how relevant is that behaviour in any other context?

• what more does 'total reputation' tell you about a person in a given context than what you can see of their behaviour in that context?

As we observed in the programme on Rethinking Personal Data, the significance and value of personal data can't be captured in a single dollar amount, or a 'yes'/'no' answer to whether it can be used. Instead, the value and utility of personal data is a hugely complex dynamic that varies by: 

• the context or the activity we are engaged in;
• which persona we are using at that moment;
• the actual data being used or provided;
• the permissions given;
• the rights that flow from those permissions; and 
• the various parties involved.

It follows that a reputation derived from a specific activity is also purely contextual, and attempts to rely on a 'good' reputation in one context as suggesting good behaviour in another are flawed. At best, as Mark Hookey conceded, the total profile or reputation data might indicate probable behaviour in another context to a greater or lesser degree, but it won't be predictive. And the person relying on the reputational data still has to know or discover the reliability of making the association.

Of course, we already know how unreliable a reputation from one context can be in a different context. Brands are key reputational badges, and while sticking a trusted brand from one industry on a new product in another market or industry might work from time to time, generally it's not a sure-fire thing. If the brand is extended to enough products that fail, the brand eventually becomes diluted, or less trusted, as the failures outweigh the power derived from success in the original context.

Indeed, I believe that internet technology is liberating us from the tyranny of a single reputation, such as a credit score.

The highly contextual nature of both identity and the behavioural data generated suggests that if you want a good reputation for doing something, then you simply need to do it and do it well. Other people will only rate you highly if you do things they find helpful (assuming you can't simply buy ratings). In other words, the vast array of reputational data available on the internet is enabling us to distinguish the facilitators, who solve other people's problems in a specific context or market, from the 'institutions' who merely claim they're here to help, but actually exist to solve their own problems at other people's expense.

So, no, reputation is not really portable. And the idea that disparate reputations can be unified or expressed as a total amount of 'reputation capital' that can be reliably leveraged over time, regardless of context, is similarly flawed.

Image from MasCanc.