In the last few days of the holiday season, we are
still in game playing mode. Our social science friends like to test
collaborative aspects of solving complex problems. Often computers are involved, rather than people, as they do what their told. That may not always be a good
thing. Real people sometimes do the unexpected.
The
latest game to be reported on is Wildcat
Wells, which Mason and Watts used to look at how networks worked in
exploring a virtual desert for hidden oilfields (1). The aim was to find as
much oil as possible in the game period. Players could drill close to neighbors
or move away and explore new regions. So getting the most oil out of the desert
meant minimizing dead wells and maximizing long term output – not too many
wells in one field.
The participants were
recruited via a web crowdsourcing program (Amazon’s Mechanical Turk) and were
assembled into networks of one of eight topologies which varied the contact
efficiency, that is some networks were more efficient that others. 232 games
were played and most players only played a few games.
Efficient networks were
found to give the best results. If a field was found, it could be quickly
exploited. An efficient network meant less copying so giving a better chance of
bringing in “the big one”.
Real players were more adventurous than computers (or their programmers) in
exploring further afield to find “the big one” compared to simulations run by
previous investigators. Search strategies were also variable with real players,
while simulations had formal rules.
It seems that good
information flow and encouraging human ingenuity is still king in problem
solving. Yeeha! we're not mushrooms.
Please note that this
blog is migrating to
in 7 days.
