Depends on the project. For most projects I would personally seek out the pragmatists in the field, as well as people with a healthy skepticism of AI as a whole. Too many AI ethics boards today whinge and fold their hands over hypothetical scenarios that aren't researched and don't pose a demonstrable harm. I would much rather identify problems that are a poor fit for AI in the first place (eg. risking human lives in a self-driving car) and avoid those, rather than pay a bunch of yes-men to feed me lines about safety metrics and diversity targets.