Launching an entirely new research area into machine behaviour as suggested by MIT Media Lab seems like an obvious good idea. Because the more we leave to machines, the better we need to understand the decisions those machines make, the rationale behind them and the impact they will have on our outcomes.
Forcing ourselves to understand machine behavior may also be the best backstop we have towards making sure that machines don’t completely take over in a ‘Terminator’-like scenario. Because even if we agree we should never get to that point, I am not overconfident that that isn’t exactly what could be happening a few decades from now (without necessarily resulting in the dystopian scenarios, Hollywood likes to present on the big screen, though).
It would also inject some much needed ‘softer’ fields of study into the world of engineering and computing, which I think we need. Not so much to keep things in check as to make sure that we really utilize technology to help us solving really big problems with a massive impact. While we, humans, remain firmly in control.