Disagree. IRV introduces some bizarre nonlinearities in the voting space that could lead to shockingly unintuitive behaviour. See these voting simulations for an example:

http://zesty.ca/voting/sim/

A better system would be something like approval voting, or condorcet, both described in the above link.

Sadly most Americans will fail to understand why IRV would be good or even how it works.

Case in point: http://www.fairvote.org/lessons-from-burlington