Tuesday, March 30, 2010

Experiment Confirms Biased Decision-Making; Irritation

"Our fears were confirmed."
It's not a statement that fills me with epistemological confidence.

I'm indeed interested in the long-term effects of short-term emotions, but the experiment design here doesn't help me. I strongly suspect that emotions have effects that last beyond their qualia, and I think ignoring or lacking this suspicion is what ruined this experiment.

The fear in question is that a bad decision made in the heat of the moment will echo in the future, and indeed I agree that bad decisions are habit-forming, as are all decisions.

The problem is that replicating the situation so closely is likely to consciously remind the subject of their bad mood. If I've been in a lab and played the economic ultimatum game once, and I was pretty annoyed at the time, no matter how long you wait, the second time I'm in a lab playing the ultimatum game, I'm going to remember how annoyed I was the first time around. Rightly or wrongly, I'm going to associate the annoyance with the game.

Andrade and Ariely did not test what they thought they tested.

"They were tapping the memory of the decisions they had made earlier, when they were responding under the influence of feeling annoyed."

False. They were being annoyed a second time around. They did not make a second bad decision while cool-headed.

"If you don’t, you may regret it. Many times over."

If indeed I'm saving time by re-using the result of a previous analysis, how important is it really if one of those times I have lost my cool? If I'm doing something for the thousandth time, instead of the first, and I happen to make a suboptimal decision due to stress, how likely is it that the thousandth-and-first time I'm going to reference only the bad decision?

For this particular example, I can't confirm I'm the best model, but I would actually improve my overall decision making as a result of this bad decision. Seeing first-hand what a bad decision looks like, I would be able to contrast the results directly to both my previous decisions, and my current decision. I would be able to recognize more subtle versions of stress effects on my decisions, because I would have essentially seen in magnified.

However, experiments cannot fail. You always learn something - even if it's just a way not to design experiments. In this case, it supports the idea that the all-important first impression is affected by 'rationally' irrelevant factors, such as overall mood that day.

I already knew this, but independent corroboration is always nice. When trying something for the first time, I start by making sure I'm in a good mood, and abort if something upsetting happens on the way there, unless I'm completely okay with a distorted assessment.

Psych. My conclusion makes the same mistake as Andrade and Ariely; the actual experiment is ambiguous. This experiment does falsify certain things, but not anything Andrade, Ariely, or I actually believe. As such, the only real conclusion is that you shouldn't design an experiment this way, and if we're going to have tax-funded science, it should at least spend some dough on experiments to pick apart why exactly this design does not work.