Big Data and Cloud Culture - To the Cloud: Big Data in a Turbulent World - page 202

unless this is clearly understood, it would be easy to extend the myth-
making about the end of the scientiic method and the end of theory and
apply it to a putative end of history, or at least of historical research, as
we have known it. This is especially tempting when the major source of
funding for historical research is a government program to make history
an arm of the digital humanities. Nor is it just a matter of taking large
data sets and putting them in a historical context. Context and history are
not discrete containers into which one can objectively insert data. They
are luid and require the experienced judgment of skilled professionals
whose subjectivity is an asset that enriches what we know, not a liability
to be set aside.
“At its core,” according to two of its leading promoters, “big data is
about predictions ” (Mayer-Schönberger and Cukier 2013, 11; italics mine).
It is hard to disagree with this conclusion and with the fact that it under-
scores both the promise and the danger of relying on large data sets. The
ability to move beyond the random sample to the billions of data points
that Google used to make predictions about the spread of the lu virus is
certainly attractive and, for some, compelling and revolutionary. But keep
in mind that even this project appears to have had a short predictive shelf
life. After a few years of success, the Google model fell lat on its face in the
2012-2013 lu season, grossly overestimating the number of cases. It is hard
to say precisely why this happened, but analysts point to the expansion of
news-media coverage of the virus's spread in December and January, which
led to far more Google searches using flu-related search terms than the
company's algorithm expected. In addition, the spike in coverage took place
during the holidays, when people have more time for both old and new
media. It appears that people were searching more not because they had lu
symptoms, but because the media stepped up its lu coverage at a time when
people were paying more attention to media. Whatever the cause, the dam-
age was done. As Google wiped the egg from its corporate face, it promised
to improve its algorithm to make better predictions in the future (Butler
2013; Poe 2013). That a similar model was used for stock-market forecast-
ing should cause concern about the consequences of overconidence in big
data for the economy (Waters 2013b). Nevertheless, economists are coni-
dent, to the point of exuberance, that big data will transform research and
policy making (Einav and Levin 2013). One of the reasons for this enthu-
siasm is the potential analysts anticipate for using big data to better manage
temporary, low-wage labor. As one report summed up, “It is rearranging