It might give the same answer, but I'm not concluding much from the thought-experiment. The model takes as input its environment, which includes its past decisions. If you pose the same question to it twice, it might choose to do something different the second time just because it weighs "whimsy" as a positive. How random is its RNG? Can it tell? What would it do differently if it knew the answer to these questions?