Interesting post by @janleike, which proposes certain desiderata for AI value alignment. An inclusive, fair, representative, legitimate and adaptive process would certainly be a good place to start👌