Doing Things with Words: Rethinking Theory of Mind Simulation in Large Language Models