Peace and quiet

Is your data really as complete as you think it is?

Today is one of those quieter days, which generates rather mixed feelings, on one hand I can get on with a tonne of stuff and finish off this nightmarish proposal but on the other hand the reason the place is quieter is that my daughter has gone back home to her mother’s and the place feels a little too quiet.

Mind you trying to work with the rolling coverage of the cruise ship disaster is difficult enough. It’s really rather incredible that such a massive vessel can be sunk by poor data. The rock that sliced that thirty metre hole in the side of the ship “wasn’t marked on any charts” according to the captain. Now whatever weight you place on his testimony at the moment, we do move billion dollar machines around the place based on the data expressed as maps and charts and as our interfaces to that data get more remove via GPS and auto-pilots our trust in the data, or rather our reliance on the accuracy of the data, becomes ever more paramount.  What gets worse, we stop trusting our own instincts when they conflict with the data presented.

Of course, what we’ve done is moved the risk point. In olden days when you moved a massive ship around the place, the navigator, helmsman (and it probably would have been a man) and captain would have combined years of experience and a keen eye/situational awareness with the undoubtedly poor data of their charts to steer a course, and of course ships sank. Now the navigator reads a computer screen and plots their course (Southampton to Rio de Janerio…click! Hell I could do that), the helmsman (and it’s still probably a man) is insulated from the rudder via a series of sail by wire interfaces and the captain is dulled by routine and instead of looking out of the window will have first consulted his computer screen to see what’s happening. And ships still sink. Perhaps more rarely but now instead of a SNAFU on board, attempts will be made to blame the data, blame the charts, “the rock wasn’t marked” etc. It gets worse though, because ships don’t sink as often the mitigation processes we developed in these cases aren’t practiced as often, people don’t know how to fit a lifejacket, the lifeboats don’t swing out and the crew doesn’t know how to react. So when things do go wrong now, they go wrong faster and get worse than before.

It’s the difference between specific risk and systemic risk and data is driving that movement, that trillion dollar movement, of risk. From sailing to derivatives data is driving systemic risk by reducing specific risk.

Why?

Simply put, people use their data to develop approaches to the risks they see, they metricise, quantatise and analyse their risks and act in such a way to reduce them, but often the mitigation actions are too often in the “transfer” bracket rather than “treat, tolerate or terminate”, and the reason is too obvious, transfer is the cheapest risk mitigation pathway. From insurance  to “we’ll use GPS instead of a 40yr veteran captain” transferring the risk is easier than fixing problems or designing the risk out.

As data becomes ever more available and as analytics becomes mainstream/commoditised and as CPU cycles become cheaper and cheaper, this movement, this trend will become more and more prevalent. Over the weekend we’ve lost a magnificent ship and perhaps dozens of lives. Over the last four years we’ve lost trillions of dollars and millions of people out of work. The data was good, the analysis was good, the interpretation was awful and the real risk appreciation was basically nil.

What happens when we make decisions on other data sets? Nuclear safety? Public health? Farming policy? Climate change?

What are we actually mitigating, where have we moved the risk to and most importantly where are the unmarked rocks? 

Anyway, back to the grind stone, let’s see if I can get questions 11-14 finished before I go out for the day and RIP for those who didn’t get off the Costa Concordia on Friday.