Theory of Knowledge - Course Companion for Students Marija Uzunova Dang Arvin Singh Uzunov Dang

Hersh, professor of political science at Tufts
Eitan
and author of Hacking the Electorate
University
idea that some additional piece of information in
The
overwhelming wave of data going into people’s
this
is going to trick them … It doesn’t give people
heads
credit.
enough
Nix’s exclamation at the beginning
Recall
this section: “we are thrilled that our
of
approach” helped Trump win.
revolutionary
Nix failed to mention was how surprised
What
team was of the result: “[a] day earlier,
his
Analytica executives told reporters
Cambridge
thought Trump’s likelihood of winning was
they
20 per cent” (Ahmed 2018).
at
himself, when interviewed on BBC
Kogan
4 in March 2018, said that the accuracy
Radio
the data he harvested had been extremely
of
He estimated that, in practice,
exaggerated.
and his team were six times more likely to
he
inaccurate information about a person’s
get
and likes and dislikes as they were to
personality
accurate information. In conclusion, Kogan
get
that microtargeting was not necessarily
thought
most effective way to use such data sets
the
are a few good reasons to be sceptical.
There
data harvested from social media, even
First,
data, does not necessarily provide
personality
actionable information or insight.
additional
other publicly available data points can
Many
a person’s political stance, including
suggest
address. Hersh states that while personality
their
are correlated with political values, the
traits
is generally weak; and that people
correlation
wrongly receive advertisements (such as
who
intended for a different demographic)
those
do not like them. For example, when he
really
to create a microtargeting model
attempted
identified people interested in climate
that
he found the best proxy is simply party
change,
if you don’t know that, everything
affiliation;
very difficult, and if you do, everything
is
doesn’t really matter. Hersh argues that
else
is effective is mobilizing voters through
what
targeting, rather than persuading
behavioural
second limitation is that almost all
The
data is self-reported, which
psychographic
it vulnerable to individuals’ blind spots
leaves
inaccurate sense of self: people who repeat
and
personality test often do not return the same
a
Additionally, their tastes and opinions—
result.
what they like on Facebook—may change,
and
they do not often go back to unlike things,
but
this sort of behavioural data needs to be fresh.
so
and most fundamentally, what does it
Finally,
to say one can “infer” political values from
mean
set of personality traits? Can we assume that
a
if the personality data were accurate, it would
Even
be difficult for microtargeting to compete with
still
information sources in the cluttered and
other
online environment. The total amount
fast-moving
political content online is so large that it dwarfs
of
output of manipulators. A team of researchers
the
2018 released results of an investigation into the
in
of infamous “Russian bots” on Twitter,
influence
looking at their ability of spreading news
When
and making it viral … we nd that their eect
content
IV. Ethics
IV. Ethics
gave the following view.
(Hersh quoted in Chen, Potenza 2018)
them to vote differently.
personalities align with politics?
with the following conclusion.
on social platforms was minor.
(Zannettou et al 2018)
55