Analytics are Key

Analytics are key, everyone agrees. And I agree too. But most times, at the start of these analytics focused efforts there is a strong consensus that, of course, these have to be 'good analytics', and not just any analytics. But by the time the conversation gets around to selecting the actual metrics to be used, we have wandered far into the territory of what is available, what is possible, what is measurable, or what is already measured, that we settle on something not just far from ideal or suboptimal, but something that is often counter-productive, or something that just perpetuates the old paradigms and a few years later, when many of the players have left the arena, we shake our head and wonder why we ever bothered.

And this is not the only pitfall on the way to using metrics and modeling effectively. It is not a simple problem to determine if an analysis or a model is good. Superficially, a model is a map that allows us to navigate a territory. So, a good one should allow more efficient navigation, getting us to our goals more quickly, more efficiently. In the same way, a good predictive framework is one that predicts future costs, health, customer satisfaction, whatever have you, more accurately than others. In either case, a good model is supposed to help us see what the most effective levers are, that could allow us to bend the curves in the direction that we want, but often our attempts to find these levers are frustrated. The reality seems to be that not only are there many levers, hidden and non-hidden, each affecting many different goals, but that these levers are interconnected, and so are the various goals, with some of them moving in opposite direction (e.g., overall health may be adversely effected by trying to raise customer satisfaction, if that means we have to allow them to see quakes whenever they want to). But in spite of knowing this intellectually, we are often unable to accept the alien logic because we want answers that are clear and unambiguous (not statistical or probabilistic), answers that are intuitive, that are common sense. But why does reality have to conform to our intuition? In fact reality is sometimes so surprising and counterintuitive that even experts deeply immersed in the complexity of the system, are heard to exclaim “Who ordered that?” (as the Nobel laureate I. I. Rabi famously did on the discovery of the Muon). Expecting complex systems to yield simple, intuitive answers can lead as on a long, unproductive, chase. Is this not the trap that Einstein fell into when he was not able to accept quantum mechanics because it was unintuitive?

How can we find a path to that utopia where all the players in the healthcare marketplace stop being focused on externalizing their own costs onto others or to the society and instead focus on the overall system effectiveness and efficiency? In response to these pitfalls of human nature and the non-intuitiveness of complex systems, I am tempted to start a list of things to keep in mind or a ‘pattern language’ of complex systems (I am certain I am not being original here, but do not know who to attribute...).

One might be that when analyzing a complex system be careful of common 'human' tendencies, tendencies that bias us in predictable ways, like:
               - assuming the shape of the solution instead of allowing the solution to  emerge from the shape of the problem (please ask me about the story of maze -solving and Connection Machines)
               - only being able to intuit linear trends
               - only being comfortable with categorical judgments, rather than where category membership is probabilistic
Another might be that whenever possible do not add another layer of complexity to solve a problem at a lower level And so on.... But, of course, by its nature, any such list has to remain incomplete.

In closing, and in response to a quote often miss-attributed to Einstein (as others do to Ben Franklin, where actually neither of them could have been its source, "Insanity is doing the same thing over and over again and expecting different results"), I offer something else, that Einstein probably did say, and is in fact, relevant to complexity and systems design: "Everything should be made as simple as possible, but not simpler".