Joshua Hartshome in Scientific American:
For years now, physicists and engineers have been building computer simulations of physics in order to understand the behavior of objects in the world. Want to see if a bridge would be stable during an earthquake? Enter it into the simulation, apply earthquake dynamics, and see what happens. Recently, the prestigious Proceedings of the National Academy of Sciences published work by MIT psychologists (and my labmates) Peter Battaglia, Jessica Hamrick, and Joshua Tenenbaum, arguing that all humans do roughly the same thing when trying to understand or make predictions about the physical world. The primary difference is that we run our simulations in our brains rather than in digital computers, but the basic algorithms are roughly equivalent. The analogy runs deep: To model human reasoning about the physical world, the researchers actually used an open-source computer game physics engine — the software that applies the laws of physics to objects in video games in order to make them interact realistically (think Angry Birds).
Battaglia and colleagues found that their video game-based computer model matches human physical reasoning far better than any previous theory. The authors asked people to make a number of predictions about the physical world: will tower of blocks stand or fall over, what direction would it fall over, and where would the block that landed the farthest away land; which object would most likely fall off of a table if the table was bumped; and so on. In each case, human judgments closely matched the prediction of the computer simulation … but not necessarily the actual world, which is where it gets interesting.