The non-intuitive result, where more memory hurts, reminds a bit of two interesting games linked off of cryptographer Wei Dai's homepage (http://www.weidai.com/), under "social costs of intelligence":
Both the seller's expected profit and the buyer's surplus are summed over two periods. The buyer has an expected surplus of 0.25 when the sale is made, but the sale is only made 50% of the time.
Interesting paper and results, but this patently does not extrapolate to "smarter parts make collective systems too stubborn."
Try: "simulated, independent agents exclusively with left/right movements collectively begin to perform worse after they are able to remember beyond 7 prior moves."
I think maybe the author extrapolated a moral opinion ("too stubborn"), and then edited out everything that supported that opinion until left with just the confusing title.
"Over the past few years, he and others have found that medium-size groups of animals or humans are optimal for decision-making"
As someone intrigued by the mechanics of politics, this really struck home. Autocracies are bad for a variety of reasons - and so too are true democracies (where everybody can vote on many issues...see the CA ballot initiative chaos in public policy).
Representative democracy strikes this "goldilocks" balance with fewer actors than full representation, but far more than a restrictive oligarchy.
It's a nice video but I don't understand how one would assign a numerical goodness measure to the starlings collective behavior. If the starlings were smarter the flock might behave differently but what would make us say that the change in behavior had a beneficial or adverse effect?
• Why good memory could be bad for you: game theoretic analysis of a monopolist with memory. http://www.weidai.com/monopoly-memory.txt
• Why cleverness could be bad for you: a game where the smarter players lose. http://www.weidai.com/smart-losers.txt