“A drawback of simulated deliberative democracy for AI alignment is the lack of clear human accountability for its decisions.”