The main question being: Did you craft a test with a wrong response? Cause I didn't, and that the core of the experiment: 80% of people proposed an answer WITHOUT having a test with a "no" response.
If you overcome the confirmation bias and tested something and get a "no" you're a better thinker than 80% of us.
I think that your conclusion is wrong. If you make several tests that yield an answer consistent with your hypothesis, you think "what else could be a valid answer, given this?" Your mind doesn't settle on "going up" because that would be very hard to test; you assume it could be something more complicated. I bet if you were to do the example again, but change the rule to "The next number must be at least 100,000/99,999 of the previous answer" the same people that guessed "it must go up" would be satisfied with the answer they received from their feedback, despite the answer being wrong. This doesn't mean they are better thinkers, it means they thought of a simpler solution and were satisfied with the feedback they received in precisely the same way as those that thought of a more complicated solution and took the feedback as confirmation. In this case the "previously right" solution would also be wrong, however. In either case, it is very, very difficult to test. In the first case (must go up) the answer might seem intuitive, and so some people will hazard onto it, but in either case sufficient testing would be extensive. This is rather an example of a very intentionally configured test that yields an intended outcome. I would bet that getting the answer right is inversely correlated with experience with mathematics.