What is the minimum number of test cases and participant groups that should be conducted for an AI to pass the Turing test?
What I mean by test cases is the number of questions that should be asked.
Participant groups - Since two humans are needed. One to Answer and one to evaluate if the response is AI or Human. How many such participant groups
Dear Cyril,
Are your questions true/false and are you trying to find a significant difference between the performance of a human and a computer? Or is the human who is checking the answers trying to discern whether the subject responding is a human or a computer based on the content of the answer?
The former is a quantitative question which also depends upon how good humans are at answering the questions. The latter is a qualitative question which depends on how good the question is at discriminating between humans and computers and how discerning the person checking is. I suppose you could also measure the accuracy of the checker's conclusions. So I think there is a lot of uncertainty about your design in its present form.
Hello Peter,
It would be the first method. A quantitative method. So I'm assuming the more tests I do the better right?
However this is my first research in this field so would appreciate any comments about it and any other alternative methods to test it.
Hello,
Perhaps G power analysis (free online tool) would be of some help
Regards, W.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude