vative. Although young women are perceived as more liberal than young men,
that gap disappears after 25.
The raw data is available at http://data.
doloreslabs.com, and O'Reilly's
Beautiful Data features the full experiment and
useful hacks for processing data.
CROWDSOURCING BENFORD’S LAW
If you asked someone to pick a number, what would it be? What's the likelihood that it would begin with one?
How about seven, or two? Benford's
Law states that about 30 percent of the
leading digits will be ones, 17. 6 percent
will be twos, and so on down to around
4. 6 percent of the numbers beginning
Because we had easy access to a
large pool of people, we tested the law
by asking subjects to pick any number
greater than zero. In a little over an
hour, we received 500 valid results that
roughly corresponded to the law, as
you can see in Figure 7. However, some
laws were made to be broken, and we
did notice a slight departure from the
law when it came to people who chose
numbers beginning with four and five.
While 42 subjects chose a number beginning with four, 51 subjects chose a
number beginning with five.
Crowds can also be used to source an-
swers to philosophical questions. Sta-
lin said, “A single death is a tragedy; a
million deaths is a statistic.” So what
about 100 deaths? What about five? We
tested this experimentally by asking
people on Amazon Mechanical Turk to
participate in the classic philosophical
conundrum “The Trolley Problem,” in
which a person must decide whether
to sacrifice one person in order to save
Figure 6: The graph above shows attractiveness based on subjects' perceived age.
Figure 5: A scatterplot matrix of the attributes: Every pair of attributes has two
graph panels. The bottom-left panels are simply x-y scatterplot with trendlines for
each attribute pair. The top-right panels show corrgrams of Pearson correlations:
blue=positively correlated attributes, and red=negatively correlated ones. Looking
at the attributes near the center, we see that conservatism, wealth, intelligence, and
trustworthiness are all positively correlated with each other, hence the dark blue
squares. Attractiveness is more complex. It tends to show a peak in the middle of the
graph, especially with intoxication, age, and political affiliation.
Figure 7: University of California-Berkeley PhD candidate Aaron Shaw
plotted this data showing the results
from 500 workers who were asked to
pick a single digit greater than zero. All
the numbers are sorted by magnitude
and leading digit. The height captures
the frequency of that leading digit. The
size of the numerals corresponds to
the magnitude of the number, and the
color corresponds roughly to its order
of magnitude (red=big). Within each
digit, the numbers are sorted along the