People Are People and Some Art

Just a jumble of thoughts today.  But first some art!

This is a barn, store front, whatever that is on the side of a road that people (like me) bike on.  The facade is normal for this area where it's built-up organically with no specific architecture in mind.  It is what it is and produces a strong character when weathered to the point it's falling down.

Done in ink and greyscale-markers.

(just my thoughts and opinion below)

I was looking at my old company's web site and discovered a couple of things.

• They are still selling and using the application I designed and wrote.  Even saw it in a video they released this summer.  It's really long in the tooth now and I tried to get it retired years ago.  Not good.
• They appear to be rehashing the concept of using "Behavior Rules" to determine risk.  It failed before but hey, sure give it another go and see how that works.

No wonder the parent company off loaded to a private equity firm recently.  Nothing succeeds like failure.

The problem isn't that behavior isn't an indicator of risk, it is.  The problem is :

• What defines what behavior is risky and how risky?  Some are easy, but there is an extremely large grey zone that could be be but mostly like not an indicator of risk.
• If we have discovered with the currently political climate, people perceive things very very differently.  Behavior and risk can be very subjective and how people think, act, operator and response to events are very hard to define. 

So given that a large set of behaviors is defined and different risk levels are assigned and possible combinations for a given timeframe within a given context produces and alert.  Simple as pie.

Sounds good, sells even better, but....

The results of this "Pre-Cog" magic is a cascade of false positives.  Risk alerts and risk levels will overwhelm analysts monitoring the output.  It's exciting for the first couple of weeks, tracking down all the specifics in the resulting data until, maybe after a week or so, they dig thought enough hay without finding any needles.   Maybe it just needs to be tweaked? 

No problem, just adjust the risk and behavior monitoring and keep going with what you have learned.

<repeat the above steps forever>

People like to make progress.  If you give a person a task like bake a cake with a recipe that will fail, they will be unhappy.  If you keep replacing the recipe with others that also produce failed results, the baker will quickly stop trying altogether.  There needs to be some path to success.

The same applies to applying behavior analysts to large numbers of people (think staff) hoping that you stumble across the magic recipe to uncover "bad-actors" is a fruitless task.  At some point the operators will stop using the product and upper management will be asking if the ROI of the product is working.

Wonderful idea but in my opinion (and from prior experience) it's doomed to fail.  The product sells but results are lacking.