Friday, March 28, 2014

Do you take a chance on someone you don't know, or assume the worst? Ask your striatum.



When and why do we cooperate?

Now that many of our systems and designs involve a social dimension, it is becoming important to understand the hows and whys of user-user interaction.  I can think of dozens of use cases for this.  Think about the social aspects of a product like FitBit. If users are going to compete or compare their times against complete strangers, some level of cooperation is needed.  If we are designing multi-player online games, players often team up to achieve quests or other in-game goals ((e.g. in World of Warcraft).  Or they might trade virtual goods in some form on on-line virtual economy (e.g. on Farmville). 

We also see user-user cooperation emerge in more basic services like commenting sections on a company’s content article (e.g. a web site’s customer support FAQs) or a news site’s article (perhaps CNN).  Or on a company’s social media page (how about this blog?).

Other use cases might include full models of crowdsourcing user-generated content (e.g. Wikipedia entries) or crowd-voting for company or user submitted materials (e.g. Threadless, Kickstarter).

Clearly, the more we know about user-user cooperation the better.  So of course I was grateful to read this article on Brain Blogger that summarizes some recent research on cooperation.  In a Prisoner’s Dilemma environment, it doesn’t make much sense to cooperate unless you think that you might develop a long term relationship with the other player and learn to mutually cooperate.  This is called direct reciprocity.  But in cases where you are only interacting with each person once, the best choice is to defect.

And yet many people don’t.  We are not utility optimizers.   Sometimes we act nice. We cooperate with this person, even though they have every incentive to defect.  But why?  That is what the research reviewed on Brainblogger was investigating.  It is called indirect reciprocity and is defined sort of like having  the notion that what goes around comes around.  They identity two kinds:

Reputation-based indirect reciprocity makes immediate sense.  If the other person has a reputation for cooperating, we can assume that they will cooperate with us, even on this first and only transaction.  So we cooperate too.  Their previous cooperation now comes around to benefit them, and our current cooperation might build us a good reputation that will come around to benefit us later.  But this only works when there are reputation signals built into the system.  This would include reviews and recommendations that are built into social media where you can report on whether your Uber driver was nice or whether your EBay seller shipped the promised product on time.

But they also found a “pay-it-forward” reciprocity.  Even if the person there are no reputation signals built in to the system, I might choose to take a shot and cooperate.  But why?  There is no possible benefit?  The other person might defect and screw me over.  And I don’t get any reputation boost from cooperating.  Why would I take this chance?  Probabilistically, this move might be good for society (if cooperation gets built into our DNA), but as an individual it is more likely to be negative.  So my economic utility maximizer (my selfish gene) should be telling me to defect.  I will win this round, and no one will ever know.

The cited research was a neuroscience study.  They used brain scans to try to figure out what parts of the brain led to this altruistic cooperative behavior.  And it turned out to emanate from the striatum.  The striatum in general is the part of the brain that processes value and executes voluntary behavior.  So reputation-based reciprocity involves the striatum asking the cognitive areas of the brain if we are maximizing economic utility by cooperating with a reputable person.  And cognitively the answer is yes, so the striatum decides to cooperate.

But when no reputation information is available, the striatum can ask the cognitive areas what to do, but gets no answer.  So instead, it asks the emotional and empathic areas.  Since we can’t go by economic utility, we have to go with what will make us feel good.  And luckily, lots of us feel good when we take a chance on somebody.  We have faith in their good behavior.  We cooperate on the hopes that they will too. 

The researchers don’t attribute it to feeling good and having faith in our fellow human.  They say it is simply a shortcut the brain is taking.  If cooperating probabilistically works (because usually we have reputational information) then we can assume that the other person will cooperate unless we have evidence otherwise.  So it isn’t faith, it is a simple overgeneralization bias of a rational utility maximizing rule. 

But I am going to attribute it to faith, hope, and charitable instincts.  I will believe in you, unless and until you give me a reason not to.