Archive for April, 2011

Joshua Greene gave a very interesting talk at Penn recently.  He discussed variants of the trolley problem, neuroscience and moral intuitions.

Trolley problems involve an out-of-control trolley that will kill 5 people who are on the track in its way, unless you do something about it.  However, doing something about it will mean killing someone else (not one of the 5 people).

Original problem:  Joe can flip a switch which will send the trolley down a different track (which happens to have a person tied to the track).    The majority of people surveyed think it’s morally acceptable to flip the switch.

Footbridge version:  Joe is on a bridge above the trolley.  Joe can stop it by pushing the large man who is in front of him off of the bridge and onto the track.  Most people surveyed do not find this morally acceptable.

Pole version:  rather than using his hands, Joe can push the man with a pole.  Most people find this unacceptable

Remote footbridge version:  Joe has a remote control some distance away from the footbridge.  It controls a trap door that could drop the man onto the track and stop the trolley.  The majority of people find this morally acceptable

Footbridge switch version:  This is the same as the remote footbridge version, except the trap door switch is on the bridge (so that Joe is in very close proximity to the trap door).  Most people found this morally acceptable.

The contribution of science is not in telling us what the right answer is in each case.  The purpose  is to isolate factors (such as spatial proximity and physical force).  If varying a factor (such as spatial proximity) causes a change in our moral intuitions, then we need to decide if that factor should matter.  If we do not think it should matter, then we have evidence that our moral intuitions are untrustworthy.

I was thinking about some new variations of the problem (hey, it’s fun to think about!).

Robot version:  there is a robot on the bridge.  Joe can use his remote control to make the robot push the man off of the bridge.  Morally acceptable? Here Joe is not using physical force, so it is like the trap door.  However, I wonder if people picture a robot that looks human (with a head, arms and legs) and associate it too closely with a human pushing the man off of the bridge.  Or perhaps it makes people think of evil, mindless robots, which makes them uncomfortable.

Lion version:  Joe is a lion trainer and his lion, Whiskers, is on the bridge.   If Joe gives the command “Whiskers, PUSH,” Whiskers  will push the man off of the bridge.  Morally acceptable?  Here Joe is not using direct force, but perhaps most people would not like it that he is making the animal do his dirty work.

Name variation:  For any of the above problems, would people respond differently if instead of Joe we called him “Damien” or “Jamal”?

Gender variation:  Would people have different moral feelings if instead of Joe we talked about Mary?  I suspect that people would be even more uncomfortable with a woman making a decision to sacrifice the bridge man.

Read Full Post »

Exaggerated effect plots

The following graph appeared in Dunbar, R. I. M. and Shultz, S. (2007). Evolution in the Social Brain. Science 317, 1344 (link to abstract).  I recommend clicking on it to see the full version.

There are many problems with the graph and the paper*, but I will focus on the presentation in the graph.  The point estimates are at the top or bottom of each box.  They also have a SE bar.  In almost all cases, if the estimate is positive, then the SE bar extends from the top of the box.  If the estimate is negative, then the SE bar extends down below the box.  Visually, this makes differences look bigger than they are.  For example, if you look at bats, the two rectangles have no overlap, and the one for pairbonded species extends all the way up past 0.4, and the one for ‘other mating strategies’ extends down past -0.3.

Compare this with my graph below, where I get rid of the boxes, and just have 95% confidence intervals (i.e., +/- 2 SEs from point estimate).  Again, I recommend clicking on the graph.

From this graph, we see a huge difference for birds, and possibly a small difference for other groups (besides primates).  But overall, it looks like birds are the only group that we can really say anything about (assuming the bird study was carefully done).


*Other things wrong with the graph & paper:

1.  where did these data come from?

2. which species do they consider pairbonded in each group?

3.  How much data from each type of species? How was this decided?  You could get small confidence intervals by obtaining a lot of data from one species in a given group.  That might not tell us much about the group as a whole.

4. What methodology was used to pool data from multiple species within a group? This is a complex problem.  It’s unclear what the population parameter is here.

Read Full Post »