But what's doing the predicting? Whatever it is, all it's doing is implementing a version of a free willed conscious being. That being has free choice, the only way to figure out what its decision will be is to progress the simulation and perk ahead. If you just keep repeating everything, that being had free choice each time and chose to do the same thing (regardless if it knew it is was in a simulation or more likely that they didnt know), in my view anyway.

If you know enough about someone that you can perfectly predict what theyll do, theres only one way to do that. Make a high fidelity copy of them and run out the simulation - which implies that we all walk around with varying quality copies of loved ones, friends, that we use when considering a gift or what they would have thought and I fully accept that and think it's true (my little strange loop copy of my partner for example isnt very good because I frequently get their order wrong for coffee and stuff). If you mean we can predict human behaviour in the aggregate, yeah I agree, but that doesnt have anything to do with individual free choice. It just means we can generally predict the activity of large enough populations of free actors.

You dont have to accept my premise that consciousness isnt something immaterial or special and that it is a purely mechanistic thing that emerges as a function or a state machine, but if you do then how that function is implemented shouldnt really be affected by if it's being run on biological hardware, in someone else's head, silicon, or being hand cranked out page after page by a room of graduate students. And in every case, there was indeed a conscious being that made free choice. At least that's what I think.