self
To be truly powerful, digital clairvoyance
will have to predict the future in specific ways
about us as individuals. This is what makes
this kind of technology categorically different
from the sort of crystal-ball fortune-telling
the world has known to this point.
“What someone like [famous US
statistician] Nate Silver is doing, in predicting
election results, is forecasting,” says Dr
Eric Siegel, author of Predictive Analytics.
“Forecasting is an overall aggregate
prediction. It’s asking, who’s going to win the
election? Who’s going to win in a particular
state across all voters?” But predictive
technology is different. “It tells you which
individual is going to vote for your candidate,
or which individual can be persuaded.”
These techniques were probably used
widely for the first time, he says, during
the Obama 2012 US election campaign,
which employed more than 50 analytics
experts, led by chief data scientist Rayid
Ghani. “They tried to work out who was
persuadable,” says Siegel. “Who would
benefit from a knock on the door?” To do
this, they built models of people, using over
80 separate information streams, including
demographics, voting history and what
magazines they subscribed to. Some of
these people received a visit from an Obama
campaigner, some didn’t. A few weeks later,
they polled them all, logging which types
of individual were swayed by the visit, and
which were unmoved or actively put off.
Now that they had rich profiles of the kinds
of humans who could be persuaded, they
could ‘microtarget’ others who matched
these profiles. “The campaign reported
a significant improvement in votes.”
Similar techniques were also believed
to be used in the 2016 Hillary Clinton
campaign. They weren’t, of course, sufficiently
sophisticated to save her from being Trumped
at the polls. But not only will their
effectiveness undoubtedly become improved,
the sheer fact of their existence represents the
arrival of a new paradigm. This is so-called
‘big data’ being used to predict which
‘individual minds’ can be changed, in ways
that already threaten to alter the course of
elections and, therefore, history.
Indeed, scientists have already found that
analyses of online data can predict historical
events. Studies of search terms made on
Google Trends, for example, showed early
warning signs of the 2008 Global Financial
Crisis. More recently, researchers including
Professor Bollen, have shown that readings
of ‘public sentiment’ can predict movements
in the stock markets.
“We looked at public sentiment as it was
gauged from very large Twitter data,” Prof
Bollen tells GQ. “Each tweet was subjected
to a language processing algorithm that
would estimate the general mood-state of
the individual who wrote it. Aggregating
that data across very large populations, we
were able to predict fluctuations in the
stock market two or three days out.”
Perhaps the company most associated with
predictions on the individual scale is Google.
Through its internet search business, its
mapping tech and its Android mobile phone
platform, the company has the ability to
model individuals (and therefore predict
behaviour) at a level of detail that, until
recently, was utterly unimaginable. If we’ve
been protected, so far, it’s been because much
of this information has been isolated. But, in
2016, the company managed to get awaywith a significant alteration to its terms and
conditions with barely a flutter of serious
media complaint. One single-line rewrite
enabled the company to combine private
account information (name, date of birth,
IP address, location, search history, emails,
contacts etc) with browsing data on apps and
third-party sites that Google tracks through
its advertising network. “This allowed the
company to create ‘super profiles’ that catalog
a user’s behavior,” offers Daniel Stevens of
the Google Transparency Project. “Google
promised it would never do this. Now, it has.”
And it seems Google’s ambitions, in the
predictive space, are soon to become potentially
even more concerning. Patents provided to
GQ by activist researchers who requested
anonymity include an application for an
algorithm that can determine a user’s mood
from a “plurality of data sources”; another that
uses facial recognition to determine a user’s
emotional reaction to a streaming video;
a possible Google ‘Home’ product that ‘learns’‘Big data’ is
being used to
predict which
‘individual
minds’ can
be changed,
in ways that
threaten the
course of
history.