IEEE Technology and Society Magazine - December 2015 - 65

for filling up on cheap food. People
who are the busiest may not be the
most productive, because they fill
their time with low-quality insignificant activities. That is because
high-quality activities take planning, organization, and money, and
when the distinction is not obvious, the incentive is to fill one's
time with the easy but insignificant activities. Those who read the
most and watch the most television
are not always the best informed,
because they read and watch low
effort and low-quality books and
programs. High-quality books are
hard to read, and when the distinction is not obvious, the incentives
are to entertain your family with
the easy-to-read books and easy to
watch programs about celebrities
and sports, not complex social and
philosophical issues.
Consider academic research.
Academic articles all look very
similar in style and presentation,
but high-quality articles are much
more costly to produce. If the quality is difficult to ascertain, then the
likely outcome is large numbers of
low-quality articles, and a proliferation of low-quality fields. The peer
review system, although revered by
the universities, does not solve this
problem, because the peers of lowquality producers are likely to be
low-quality producers themselves,
and may not even recognize highquality work. More importantly,
peers have no incentive to pay the
high price of evaluating quality in
terms of time and effort. This may
explain the proliferation of new
fields of study, and the increasing
numbers of Ph.D.'s and academic
publications, where large quantity
invariably chases out high quality.
Harvard economist John Kenneth
Galbraith once lamented that the
shifting emphasis in economic
research from qualitative analysis to
empirical data analysis had impoverdecember 2015

ished the field of economics. That is
the conclusions, and almost never
because the emphasis had shifted
ends a debate conclusively. Morefrom domain knowledge and insight
over, one can potentially find some
to learning and properly using statisstatistical evidence to support any
tical tools. The qualitative research
hypothesis, because 95% significould not compete with empirical
cance test is rather weak when so
work, not because it was lower qualmany researchers are repeating the
ity, but because empirical work was
same tests again and again with difmore easily produced and evaluferent data sets, and ignoring the
ated and hence less risky.
Ph.D. students could count
on finishing on time, and
faculty could count number
of publications for evaluaFrench poet Paul Valery
tion purposes. Qualitative
work was risky, subjective
wrote in 1895: "Western cultures
to evaluate, and difficult to
worship information as if it
predict success. As a result,
empirical work crowded out
were an omnipotent beast and
qualitative work, and academic evaluations were
place no limits on what they
reduced to counting the
seek to know."
number of publications per
year [11].
Similarly, UCLA statistician Judea Pearl criticizes social science research as
failures, and publishing the rare
overly concerned with statistical
successes. The inability to replicate
tools, and not enough with insight
most research findings is a troubling
and policy recommendations. In
consequence of this research envifact, he argued that the empharonment. As a result, the explosion
sis on statistical tools made social
of research activity does not provide
science research less relevant to
better insight into any of the hypothunderstanding and solving real
eses tested. This is how quantity
social problems. In every empirical
overcrowds quality even in scientific
research paper he colorfully proresearch [19].
nounced that there is one section
Consider a highly publicized
that is completely nonscientific and
research article by Cornell Uniit is titled conclusions, and that is
versity psychologist Daryl Bem in
where he authors often draw policy
which he found statistically signifiimplications. But policy implications
cant evidence that humans can prerequire an understanding of caudict purely random future events
sality so that one can manipulate
[5]. Such clairvoyance contradicts
causes to change effects. Yet, stathe laws of physics as we know
tistical data analysis without experithem. Consequently, this research
ments rarely provides insight into
presents a dilemma to either reject
causality, but only correlation, and
the laws of physics governing time
the policy conclusions drawn are
and randomness, or reject the
merely opinions with little scienrules of statistical significance.
tific basis. That is why most every
Either option requires invalidating
research paper ends with an urgent
a large amount of existing human
plea to do more research to justify
knowledge. There is evidence that
∕

IEEE Technology and Society Magazine

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - December 2015