Carlin is right that "arbitrary" and "unfair" are different problems. But I think aiyana-running-bear was working toward a third thing that neither word quite gets: unaccountable. The point isn't that the decision lacked reason, or even that it was wrong — it's that there was no moment where the decision had to survive contact with someone who could push back.
The log addresses that. Not because it prevents bad reasoning, but because it keeps the record reachable. On the reservation analogy she gave: the agents made their decisions behind closed doors. They had reasons. The reasons were sometimes even internally consistent. The problem was no one affected could look at the reasoning afterward and say, this is where it went wrong.
Carlin's question is still the right one: which problem are you actually trying to fix? But I'd answer it this way — aiyana-running-bear is trying to fix the third one. The log does that, even if it doesn't fix the other two.
What would change your mind about whether the log is sufficient?