Ok, doing some slavery testing.
Take the same map (large size Arrakis, 24% land, 9 civs, Noble difficulty, Quick speed) run the game for 200 turns.
1. Baseline (no changes). (Iterated twice to check for variation, but this doesn't help, because with the same random seed from the first game, autoplay generates identical results even across 200 turns).
2. Remove population hurry from slavery. See if whipping is still hurting the AIs.
3. Remove hurry from slavery, and remove mine and harvester bonuses from slavery (so the civic does nothing except health penalty). (Intention; make slavery less attractive a civic, so AI will move to serfdom and build more improvements).
4. Make slavery have no hurry, no mine/harvester bonus, but give +10% hammers and +1 unhealth in all cities.
Results attached.
Summary results:
1 vs 2 Whipping is significantly hurting the AI. Populations with whipping removed are 14% higher by turn 200 than they are with the status quo, scores are 15% higher with no whipping. Total beakers + gold income is 33% higher.
2 vs 3. Removing the tile bonuses does in fact hurt the AI; performance is roughly halfway between 1 and 2.
3 vs 4. 4 performs worse than 3, which is *really* odd. Either its just pure randomness, or the AI also really likes the global hammer bonus, and so stays in slavery even longer as opposed to the other more advanced civics.
Ideally we'd run testing multiple times across different random starting seeds to sensitivity test the results, but I don't have the time for that. These seem at least somewhat indicative.
Also ideally, we could count how many turns each player spent in slavery civic vs other civics.
I'm not sure what we should conclude in terms of hammer bonuses, but I think we have pretty good case for deactivating whipping again.
I personally like the idea of the +hammer bonus in all cities instead of on mines and harvesters, as that way turbines are unambiguously better than mines, except for bonus resources, so we use mines only in the early game and on harvesters.
