Yes, but also, resolutions of human preference conflicts should eventually be done with some input from human meta-preferences, not solely in the simplest way.
I wonder whether this framework allows for meta-preferences to play a role. Maybe first-order logic constraints on the graph edits, or penalties for specific edits (e.g. removing edges is 1.5 times worse than adding edges…)
Yes, but also, resolutions of human preference conflicts should eventually be done with some input from human meta-preferences, not solely in the simplest way.
I wonder whether this framework allows for meta-preferences to play a role. Maybe first-order logic constraints on the graph edits, or penalties for specific edits (e.g. removing edges is 1.5 times worse than adding edges…)