Woah! I think this is a great initiative to log human behavior in complex situations. And as the OP pointed out the data could then be used to model and mimic himan decision making.
That said, the role of a regulator for AI agents cannot be over-stated. As machines start making these high impact decisions, it would be necessary for a regulatory body to spell out a broad framework that can guide the automatic decision making. However, the task of defining the specifics of such a regulatory framework will be non-trivial and would need collaboration between AI Engineers, law-makers, corporations and society at-large. And this website seems like a great way to illustrate the challenges and start a conversation.
The other thing that comes to mind, with regards to this website is the role of decision fatigue. I would be curious to see if the decisions towards the end of the experiment were consistent with those made at the outset.I wonder if the string of difficult scenarios presented on the website can lead to logging of sub-optimal decisions towards the end of the judging experiment.