[FoRK] Human behavior is 93% predictable

J. Andrew Rogers andrew at ceruleansystems.com
Thu Mar 4 20:02:41 PST 2010


On Mar 4, 2010, at 7:28 PM, Ken Ganshirt @ Yahoo wrote:
> Okay, stay with me here. You said:
> 
> "Computers will see all manner of exploitable pattern that humans do not and those abstract patterns are very difficult for humans to reason about in the context of the overall model. "
> 
> If the humans managing the computers can't see or reason about those abstract patterns how can they figure out how to exploit them? 


It's like fly-by-wire for an aerodynamically unstable aircraft. Stable flight is beyond the capability of a human but a computer can handle it just fine by simultaneously accounting for far more factors and variables and their aggregate impact on future states of the aircraft than a brain could.  

In these aircraft, a human communicates a desired action or state to the fly-by-wire computer which computes the best way to deliver that result and then executes it. The human pilot has no real visibility into the details of how the operation was achieved, but they get the result they were looking for. At the same time, the software was clearly designed by a human that understands theory of stabilizing unstable aircraft even if they can't mentally replicate the operations of the computer.

Same kind of story. We know the patterns exist in the abstract and their theoretical relation to one another. We can't discern them ourselves (a computer has to tell us they exist) nor can we properly grok the meaning of any particular pattern or apply it. However, we can easily conceptualize a desired result and instruct the computer to apply theory to achieve that result using whatever patterns it can discern, or alternatively tell us that such a result is not feasible.

There's no magic to it. A lot of computerized systems work this way.




More information about the FoRK mailing list