This applet at the New York Times website is very nice: http://www.nytimes.com/interactive/2011/04/18/science/20110419-music-expression.html.
It asks us to identify the human piece, presumably the “most expressive”, from variations that have been created in many ways. I managed to identify the human piece 75% of the time but in at least one case, the decision was a very close one and I was initially leaning towards the randomized mechanical one.
The big question of course is how to characterize the contributing factors. These kinds of tell-the-difference tests lasting less than a minute are nice to drive home a point but if we want to create the real thing without just replaying, it would be good to understand a bit deeper. I’d love to see more on this topic in the popular press, or even in the scientific press but in a way that non-musicians like me can understand!