Pen's Nest

I'm not too surprised that Facebook thought that was a cool idea -- partly because Facebook views users only as ad clicks, not people, but also because I think it's a cool idea, too. But then I'm a scientist, and we tend to think that kind of thing is cool, which is why there are ethics committees and institutional review boards who are supposed to stop us running amok.

I am rather surprised that the University of California, and Cornell thought it was appropriate, and that the PNAS published it. The paper mentions the data collection methods, but it doesn't say anything about whether the design passed an IRB. But then, this isn't the first time I've seen researchers with the attitude of 'it doesn't matter, it's only the internet'.

The paper doesn't even show what they say it shows. As they didn't actually measure mood, only use of positive and negative words, I think they're exaggerating their conclusions. The effect they've demonstrated isn't necessarily emotional contagion, it's the way the emotional environment affects what people think is appropriate to say out loud. If people make more positive or negative posts, it could be because their mood has actually been altered, but it could also be because they feel more or less comfortable about expressing those feelings. I didn't see any way they tried to differentiate or control for that.

Also, the actual effects they get are miniscule. The biggest effect they saw was that decreasing the number of positive posts seen decreased positive words in posts compared to the expected number by 0.1%. Big whoop. The effects are only significant because of the huge volume of data they could get.

Not that the feeble results make the experimental protocol any less wrong, of course.

(Link to the full article.)