Tuesday, January 31, 2012

collaborative learning: two sides of a coin

Collaboration is on my mind these days. Ingrid pointed out this article to me, about, among other things, the potential pitfalls of group work. The author, Susan Cain frames this in the presence or importance of introverts, and how negating introverts has a negative effect on the performance of group work. Choice quote from the interview:
"Forty years of research shows that brainstorming in groups is a terrible way to produce creative ideas."
(After I blogged this, I found a similar article were Susan Cain explains it in her own words: http://www.nytimes.com/2012/01/15/opinion/sunday/the-rise-of-the-new-groupthink.html)

And guess what, serendipity strikes again: I had flagged this recent article in PNAS for a closer look and a potential blog article. It actually explores this tension between individual versus group learning. The outset of the paper is the experimental result that in networks of learners (i.e., collaborators), reducing the speed of information exchange between learners (or network efficiency) actually leads to higher performance of the network as a whole. Or translated into the Cain framework: introverts can be introverts, do not have to adjust to group thinking, and can thus lead to higher performing networks (note: this is not necessarily the 40 years of research that Cain refers to). The mechanism provided in the PNAS article puts this in more scientific terms:
"[T]he explanation is that slowing down the rate at which individuals learn, either from the “organizational code” (3) or from each other (8, 11, 13), forces them to undertake more of their own exploration, which, in turn, reduces the likelihood that the collective will converge prematurely on a suboptimal solution."
The problem with this mechanism is that it is mainly based on computer simulations. The novelty of the article is that they tested it with real, human, players using a large sample size. And they found the opposite results:
  • network efficiency was positively correlated with performance
Average points earned by players in the different networks over rounds (error bars are ±1 SE) in sessions where the peak is found. Graphs with high clustering and long path lengths are shown in dark gray; those with low clustering and low path lengths are shown in light gray.
  • players copied less in more efficient networks
(A) In contrast to theoretical expectations, less efficient networks displayed a higher tendency to copy; hence, they explored less than more efficient networks [numbers and colors (orange is shorter and green is longer) both indicate clustering coefficient]. (B) Probability of finding the peak is not reliably different between efficient and inefficient networks.
These results indicate that in networks with efficient information exchange, individual behaviour (of for instance introverts) is valued, and this individual behaviour helps to increase the performance of the group. Or the apparent contradiction between collaboration and introverts is saved!

This blog post provides some real-world evidence for this. It is a book review of Steven Johnsons's Where good ideas come from. John Batelle extracts this nice figure (which of course appeals to myself as a scientist), illustrating that major ideas arise more frequently in a networked environment:
Johnson's chart of major ideas emerging during the 19th century, categorized by commercial and networked properties

The two sides of the collaboration coin are thus individuality and efficient information exchange.

It is not that simple, of course, but at least it provides a mechanism to for instance increase the efficiency of collaborations in a class room setting:
  • it is important that everyone participates
  • it is important that everyone thinks about the issue by themselves, before sharing it with the group
  • it is important to listen to everyone
  • the flow of information between participants should be as efficient as possible


  1. thanks for posting this. very interesting - as the other texts you refer to.

  2. If we make the parralel with metacommunity or network theories, these empirical results are quite countertuitive. I would expect, as the computer models showed, that when information's dispersal (or connectance) is too high it reduces the overall functioning of the network. Very interesting!

    1. It depends. This would mean that information would result in mass effects, or that an individual would disregard his own data on the success of a certain strategy, based on seeing a neighbour repeatedly trying a suboptimal solution. In the game that they were playing, that would be very unlikely. However, in real life without such a clear optimal solution based on sampling, this mass effect is very likely. You could for instance explain the success of Fox News as a mass effect.