Systemic bias is the inherent tendency of a
process to favor particular outcomes. The term is a
neologism that generally refers to human systems;
the analogous problem in non-human systems (such as measurement
instruments or
mathematical
models used to estimate physical quantities) is often called
systematic bias, and leads to
systematic error in measurements or
estimates.
Bias in human institutions
One might refer, for example, to the systemic, systematic, or
institutional bias of a particular institution in devaluing
contributions by women, men or ethnic minorities. For example, a
poetry competition that was consistently won by white women could
be subject to suspicion of a bias if there were no inherent reason
that white women would consistently be the best poets. Such a bias
could be deliberate on the part of the judges or entirely
unconscious.
For example, the poetry contest might be judged by a pool drawn
from its own previous winners, reasoning that prize-winning poets
are the best to judge a poetry contest. However, it might be that
in addition to choosing for poetic skill, they are also inclined to
choose people with whom they have values in common, either about
poetry or about other matters, resulting in a continuous stream of
prizewinning white female poets. In this case, the bias could arise
from either conscious or unconscious defense of
gender and
racial interests or
simply from their shared
point
of view. In either case, it results in a biased representation
of the reality they are describing in terms of quality of poets and
poetry.
Because
cognitive bias is inherent in
the experiences, loyalties, and relationships of people in their
daily lives, it cannot be eliminated by education or training, but
awareness of biases can be enhanced,
allowing for the adoption of compensating correction mechanisms.
For
example, the theory behind affirmative action in the United States
is precisely to counter biases in matters of
gender, race, and ethnicity, by opening up
institutional participation to people with a wider range of
backgrounds, and hence presumably a wider range of points of
view. In India
the system
of scheduled castes and
tribes was intended to address systemic bias within the
caste system. Similar to affirmative
action, it mandates the hiring of persons within certain designated
groups. However, in both instances (as well as numerous others),
many people claim that a reverse systemic bias now exists.
Examples
Financial Week
reported May 5, 2008 (
emphasis
added):
But we travel in a world with a systemic
bias to optimism that typically chooses to avoid the topic
of the impending bursting of investment bubbles. Collectively, this
is done for career or business reasons. As discussed many times in
the investment business, pessimism or realism in the face of
probable trouble is just plain bad for business and bad for
careers. What I am only slowly realizing, though, is how similar
the career risk appears to be for the Fed. It doesn’t want to move
against bubbles because Congress and business do not like it and
show their dislike in unmistakable terms. Even Federal reserve
chairmen get bullied and have their faces slapped if they stick to
their guns, which will, not surprisingly, be rare since everyone
values his career or does not want to be replaced à la Mr. Volcker. So, be as optimistic as possible, be
nice to everyone, bail everyone out and hope for the best. If all
goes well, after all, you will have a lot of grateful bailees who
will happily hire you for $300,000 a pop.
Systemic versus systematic bias
There is some contention over the choice of the word
systemic as opposed to
systematic.
"Systemic bias" and the older, more common expression "systematic
bias" are often used to refer to the same thing; some users seek to
draw a distinction between them, suggesting that systemic
bias is most frequently associated with human systems,
and related to
favouritism.
In
engineering and
computational mechanics, the word
bias is sometimes used as a synonym of
systematic error. In this case, the bias is
referred to the result of a measurement or computation, rather than
to the measurement instrument or computational method. Thus,
expressions such as "bias of a measure" are sometimes used.
Systematic bias is rarely used and
systemic bias
is never used with that meaning.
Some authors try to draw a distinction between systemic and
systematic corresponding to that between unplanned and planned, or
to that between arising from the characteristics of a system and
from an individual flaw. In a less formal sense,
systemic
biases are sometimes said to arise from the nature of the
interworkings of the system, whereas
systematic biases
stem from a concerted effort to favor certain outcomes. Consider
the difference between affirmative action (systematic) compared to
racism and caste (systemic).
See also
References
- "Paging Paul Volcker. The former Fed chairman was
tougher and less eager to please than his successor, Alan
Greenspan". By Jeremy Grantham. May 5, 2008. Financial Week.
Further reading
Wikis
Quiz
Map
Search