Thursday, 25 September 2014

Wearables: Three issues for Insurance to try on for size

A supplement in The Times newspaper has quoted from my analysis of the potential of wearables to change the way the insurance industry works. I've published the analysis below in full and the Times article is here.

Wearables + the IoT = Lots of excitement
Wearable technology becomes exponentially more interesting for the insurance industry when the data that devices collect about our behaviour are combined with data emanating from the broader ‘Internet of Things’ (IoT). In fact insurance has been called the‘nativebusiness model’ for the IoT, a term that describes the anticipated 26 billion internet connected objects by 2020, in the same way that advertising was dubbed the native business model of the internet. 

The Internet of Things. That's the dry theory but how about messy reality?
Phil Windley ‘Web ofThings’ CC BY-NC-SA 2.0 licence 

With more data to cross reference, for example data collected about our exercise regime through a wearable health bracelet with what smart cutlery says we are shovelling into our mouths, the theory is that predictions can be made more accurate and premiums more reflective of our lifestyles.

Steady on! Three issues to address
Before the insurance industry gets too excited about the potential of wearable technology and the IoT however, there are three crucial issues to consider. Firstly, whether wearables become ubiquitous is out of the industry's hands and depends on the extent that they enhance our lives as social beings. Secondly, the industry needs to find ways to access available sensor data and finally if it does, it may find that too much exposure to our personal data exhausts may be as useless as too little. In more detail:

 One - Factors affecting the popularity of (particular) wearable devices
Wearable technology is not new and has social meaning
One Lucky Guy ‘Knight of the Hundred Years War’ CC BY-NC-SA 2.0 licence

The main challenge for the manufacturers of wearable technology is how to make their products desirable. Their success will dictate the uptake of wearables and therefore their usefulness to the insurance industry. The most forward thinking technology companies have realised both that what we wear has symbolic value, plays into cultural processes and that data alone is of limited use to us as social beings. Why else would there be a solid gold version of the new Apple Watch, which has self tracking functionality, other than to signal something about ourselves? Being interested in data per se is a niche pursuit an even then has social aspects, as demonstrated by the practices of 'quantified self' movement adherents. For others the connected nature of the new generation of wearables can help them play into age old inter-social dynamics such as bonding and building our reputations. 

Worn in the same place and about fashion and so many things besides
dlane cordell ‘Time to get fit 5/09/14’ CC BY-NC-ND 2.0 licence

The success of particular wearables is not guaranteed (look at what happened to Nike's ostensibly successful Fuelband) and the insurance industry will have to understand which have the best prospects based on these criteria, before getting into bed with them, or rather slipping them onto their wrists.

Two - Accessing the data
Accessing the data is the next issue for the industry. Privacy concerns, made all the more real by Snowden’s NSA revelations, will potentially temper peoples’ and societies’ willingness to share the data they generate voluntarily or inadvertently with companies and by extension the government. Further, insurance companies are not the natural data gatekeepers. They will either have to work with those who are, for example Microsoft’s deal with American Family Insurance earlier this year to find ways to put sensors into our domestic environments, or let Google, which already knows far more about our risk profiles through its plethora of platforms, dictate the terms.

Three - More data brings more headaches
Finally, big data analytics is in its infancy and struggles with the same problems that statisticians always have done. Leaving aside issues of the compatibility of different sources of data, one of the most relevant issues for the insurance industry that may suddenly have access to many more variables from our personal data exhausts, is how to work out what patterns are significant. This problem is compounded by very real disagreements about the virtues or otherwise of certain practices: It's all well and good being able to track that an individual has just had a glass of red wine but the medical establishment itself can't agree on whether that's a good or a bad thing.

Two rather more interesting questions
As a postscript we could ask two perhaps more interesting questions. How does society stand to be changed by the interest of the insurance industry in collecting more behavioural data and why do we behave in the way such data show we do?

One - Are we coming to assume you have something to hide?
Andrea R ‘Sarah has nothingto hide’ CC BY-NC-SA 2.0 licence 
By rewarding or insisting on transparency the insurance industry reinforces the assumption, espoused by governments and tech giants such as Google that we have something to hide if we don’t want our lives to be subject to close examination. But there are a number of very good reasons why the prospect of us becoming ‘entrepreneurs of the self’, namely managing personal data portfolios that we selectively release in exchange for perks, is not adequate compensation for the all-encompassing surveillance that is entailed. One of those reasons has to do with the use to which that data is put, for example by the government (which has access to all of it) to make predictions about our future behaviour that create suspicion based on obscure algorithms. Incidentally, insurance companies grappling with our data exhausts will also increasingly use obscure algorithms over which we have no recourse and that stand to entrench discrimination if for example the wealthy can opt out over the kinds of surveillance and behavioural controls that the poor are subject to in order to qualify.

Two - Big data needs small data
Big data needs to be complemented by ‘small data’ (or 'thick data') collection if it is to mean anything. A series of ‘big data’ sensors in smoke alarms might be able to tell an insurance company that they are routinely left with flat batteries but until you investigate the reasons why by looking at the personal and social contexts that influence individuals’ decisions not to replace a battery, you aren’t going to have much luck in changing behaviours. That’s where a digital anthropologist comes in…

No comments:

Post a Comment