I disagree. I think it would have a very good chance to work.

To a perfect Bayesian, the importance of an act is not what it looks like on the surface, but the state of the world that makes such an act possible. Unfortunately (or fortunately in this case), human minds are not perfectly Bayesian.

To the human mind, merely resembling another thing is enough for the mind to form connections and associations between the two. This is why public speaking courses can improve people's abilities and lessen their fears of public speech. Even though people know they're just speaking in front of a class who is obligated to receive the speech well, their mind naturally reduces the anxiety they feel for any future speaking engagements. The mind says "eh, it's close enough. I can do this," just like how anti-conformity training should fool the mind into considering it 'close enough' to real disagreement. Speech classes don't not work perfectly, just like chaos training (I assume) doesn't work perfectly, but it's pretty good.

Anti-conformity training seems practically identical to a proven training method, and thus I rate it highly likely to work.