Common Sense: Limits on Testing From Turing to Self Driving Cars

Note: This is then thirteenth post in an occasional series on Common Sense Election Integrity, summarizing, updating, and expanding on many previous posts covering election integrity, focused on Connecticut. <previous>

At first this may not seem like Common Sense. We have the famous Turing Halting Problem which has some very important consequences for voting which may not, at first, make common sense:

  • We cannot use testing to be sure that the software in a voting machine will provide accurate election results.
  • And any hardware circuits are also part of the machine and come under the limits of the halting problem

It is worse, beyond the halting problem:

  • We really have no way of knowing if the software that actually ran on a machine when the results were created and printed was actually the approved, tested software.
  • We really have no way of determining if the results were somehow changed by some some means external to the software.
  • We have no way of really determining that the components of the hardware were what were tested were actually those running the machine.
  • There could also be permanent or intermittent hardware errors.
  • The hardware errors could include logic circuits, wires, or sensors.

At this point you may be complaining that this is crazy or at least not common sense.

Consider the idea of self-driving cars.  How comfortable are you with them today?  Do you think testing is sufficient?  Maybe. Yet, they could be subject to intermittent errors and hacking – similar to today’s vehicles that rely almost entirely on software to translate the driver’s commands into action. See:  <60 Minutes Shows Threats to Autos and Voting Machines are Real>


Leave a Reply

You must be logged in to post a comment.