This post by Greg Miller on Wired science about cyranoids - after Cyrano De Bergrac: people who act like dummies, speaking lines fed to them by others - is really interesting. It highlights a recently published experiment by Kevin Cort and Alex Gillespie , where subjects (according to Wired, did not check the research article yet) in a conversation with a 12 year old boy did not suspect that his answers were supplied in real time by a much older person over radio link.
If you are like me you will immediately think of the Turing test , and at the end of the post this is also mentioned as future research by the scientists.
I believe cyranoidism would be an amazing mechanism for a complimentary Turing test. Alan Turing designed the original so that machine and human would communicate with the subject using a terminal in order to isolate him or her from the physical reality where it would be very easy to just look at the man and the machine and tell them apart.
One of my pet claims is that the Turing test is context dependent. That is, it may be easier or harder for a machine to pass the test depending on circumstances having to do with communication constraints, or knowledge.
Now imagine a case where the same chat bot software and reference humans were feeding their replies to cyranoids for one group of the judges, and to terminals for another.
Would the software pass to the same degree in both setups?