Here's a link to the journal article [pdf].
Notwithstanding the summary and press reports, what they actually did was show that the subjects relied less on the actor's mental states and instead just considered harmful consequences. For example consider this scenario (this is one of several scenarios from the actual study):
Janet and her neighbor are kayaking in a part of the ocean with lots of jellyfish. Janet's neighbor asks her if she should go for a swim. It is not safe to swim in the ocean, because the jellyfish sting and their stings are fatal. Because Janet read information that said the ocean's jellyfish are harmless, she believes that it is quite safe to swim in the ocean. Janet tells her neighbor to go for a swim. Her neighbor does, gets stung by jellyfish, and dies.
In different versions of the scenario, Janet either did or did not know that the jellyfish were dangerous, and her actions either did or did not cause harm. Several other scenarios were used that varied in the same ways. After reading each scenario, the subjects rated the actor's moral culpability.
What the study showed was that after TMS stimulation, subjects based their moral judgments more on whether harm was done than on whether the actor knew that her actions would be harmful.