IEEE Technology and Society Magazine - Summer 2013 - 27

arguments include setting the
beginning of life at conception or
even earlier, determining first trimester survivability of the fetus
out of the womb, characterizing the
"morning after pill" as abortion,
whether to make exceptions for
rape and incest, deciding whether
pregnancy is common in case of
rape even though rape treatment
in every hospital includes highly
effective pregnancy prevention,
and a myriad of other tangential
issues which have nothing to do
with the political struggle. As the
positions become entrenched and
increasingly couched in moral
principles to rally the simple folk,
the rank and file, the underlying
political issues are forgotten, and a
compromise becomes impossible.
The social cost of such unending
political debates and struggles is
astronomical, especially if they
turn violent [15].
Similar debates take place about
environmental pollution. Industry
often takes the extreme position
of new chemicals being considered completely safe until there is
conclusive evidence that they are
harmful. Even with considerable
evidence to the contrary, the debaters may get locked into their positions. Jack Welch, former CEO of
General Electric, once claimed that
PCB's were completely safe, and
the pollution of the Hudson river
was exaggerated, since as a young
chemist, he was in PCB's "elbow
deep" every day, with no harmful
health effects [21]. Similarly, some
environmentalists would like to
ban all new synthetic chemicals
until they are conclusively proven
to be safe to all humans, animals,
and plants. Such extreme positions
do not lead to useful compromises;
and with thousands of new chemicals being introduced to the environment every year, it is difficult
to reach a reasonable conclusion to
this debate.
In a fast changing technological environment, with entrenched
interests constantly threatened,

it is difficult to have reasonable
compromises and make reasonable
policy decisions. Instead, unending
debates create a class of professional debaters, building a career
out of representing various interests. They tend to be ideologues
and demagogues, with no serious
interest in finding compromises,
since their own interests are in
building careers out of unending
political struggles [10], [15].

Prediction and Risk
Statistical analysis provides the
tools for studying risk and making predictions, and statistics has
also become the mainstay of discovery and analysis in social sciences. There are numerous and
fundamental problems with the
extensive reliance on statistics
to study and evaluate risk, and
to predict and remedy its consequences. First and most obvious
is that it takes attention away from
the domain knowledge and insight,
and places the emphasis on formal statistical tools. John Kenneth
Galbraith, Harvard economist and
past president of American Economic Association, once famously
lamented that statistical analysis
had impoverished economic studies, by focusing too much on quantification, at the expense of insight.
unfortunately, insight is critical
to developing solutions to social
and economic problems. Statistical analysis may find relationships
among existing phenomena, but
does not find solutions to change
the undesirable phenomena. Solutions require insight into causality,
not just relationship and correlation
among variables.
More importantly, without
causal explanations and insight,
statistical relationships are often
interpreted as suggesting the most
obvious solutions, when in fact
proper solutions can be far from
obvious. Such statistical studies
of risk ironically carry a huge risk
themselves of missing the causes,
explanations, and the consequences

IEEE TECHNOLOGY AND SOCIETY MAGAZINE

SuMMEr 2013

of risk, leading to inappropriate
responses. Judea Pearl, uCLA
statistician, once argued humorously that in every academic
paper there is one section that is
completely nonscientific, and it is
always titled "conclusions." Statistical analysis rarely gives enough
insight into causes of social issues
to allow correct policy conclusions
[11], [25].
Consider the statistical finding
that children who grow up with
two parents are healthier and more
successful in life. The immediate and obvious policy conclusion
is to support the marriage institution and to discourage divorce.
In fact, that might be the wrong
policy conclusion without further
insight. One needs to understand
why people divorce and how that
affects children, which is called
the process or domain knowledge.
For example, it is possible that if
you separate good marriage and
bad marriages, the children of bad
marriages may be better off after a
divorce. If you distinguish between
amicable divorces and hostile
divorces, the children of amicable
divorces may be healthier and
more successful than the children
of married couples. And those are
just two simple examples. There
are infinitely many such variables
that may allow fine tuning of any
statistical analysis, such that without a detailed qualitative understanding of the divorce process and
how it affects children, it would
be difficult to draw any policy
conclusion for a purely statistical
analysis. Without such insight, calculating the risk to children from
divorce would be too general to
be useful, or even fundamentally
incorrect [25].
Adding to the difficulty, statistical analysis focuses on expected
risk, not the risk of extremes and
catastrophes. Gamblers for example often calculate their odds of
winning, but the expected value
may not be very useful. Even if the
expected value was positive, most
|

Table of Contents for the Digital Edition of IEEE Technology and Society Magazine - Summer 2013