88 Zombie Media
pages against a database of 2.8 billion facts and thus influence its page
ranking in Google’s search engine.
35
Observing that most attempts
to rank pages depended on exogenous signals such as hyperlinks to a
site, indicating its popularity, the authors based their work on assessing
trustworthiness as defined by accuracy and shared some links with work
undertaken by Ray Kurzweil, Director of Engineering, to create a form
of artificial intelligence that could understand content without relying
on such third-party inputs. While the authors identified the possibility
that accuracy could be ranked (at least when dealing with the database
of facts to hand), several obstacles remained in the way of its implementation: irrelevant noise, trivial facts, and a high percentage (some
15 per cent) with regard to false positives for accuracy. In the end, the
authors concluded, algorithms would be able to determine the truth of a
page based on the accuracy of its facts, but such accuracy was nowhere
near good enough for release into the wild.
As such, and leaving aside for one moment Hume’s distinction between matters of fact and relations of ideas in An Enquiry Concerning
Human Understanding (to which truth as generally accepted reasoning
may be said to belong), this leaves us with a trade-protected page ranking system that depends on elements such as popularity as determined
by hyperlinks or aggregated searches, the exogenous links referred to
by Google’s engineers. Similarly, such aggregated data inform Google’s
autosuggest feature, helping searchers by beginning to complete their
search terms.
Carole Cadwalladr, writing in The Guardian in 2016, described her
experience of undertaking a thoroughly mundane, twenty-first-century
task: googling a search term. Entering the letters a-r-e j-e-w-s, Google
autocomplete offered a range of suggested questions: “are jews a race?”,
“are jews white?” “are jews christians?” and “are jews evil?”
36
Alongside a range of suggested answers to the questions (including that “Jews
are demonic souls from a different world”), Cadwaladr discovered that
the direct answer autocorrect suggestion to the word “women” was “are
women evil?”. In 2006, Google had been subject to a lawsuit by the
Anti-Defamation League because the first result for “jew” had been a
link to the anti-Semitic website, jewwatch.org, leading Google to emphasise that the page ranking system was automated. Over the next few
years, the processes involved in generating Google results continued to
attract attention from time to time, as when Tom Chatfield drew attention to the search engine’s disturbing trend of suggesting that women
shouldn’t work, shouldn’t vote, and shouldn’t even have rights.
37
In
general, Google had taken a lofty role with regard to its position, exonerating itself as a company with regard to its duties of overseeing its
algorithmic curators. In December 2016, however, Google was one of
a small handful of tech companies being blamed for the surprise outcome of the American election. Cadwalladr’s story was published on
pages against a database of 2.8 billion facts and thus influence its page
ranking in Google’s search engine.
35
Observing that most attempts
to rank pages depended on exogenous signals such as hyperlinks to a
site, indicating its popularity, the authors based their work on assessing
trustworthiness as defined by accuracy and shared some links with work
undertaken by Ray Kurzweil, Director of Engineering, to create a form
of artificial intelligence that could understand content without relying
on such third-party inputs. While the authors identified the possibility
that accuracy could be ranked (at least when dealing with the database
of facts to hand), several obstacles remained in the way of its implementation: irrelevant noise, trivial facts, and a high percentage (some
15 per cent) with regard to false positives for accuracy. In the end, the
authors concluded, algorithms would be able to determine the truth of a
page based on the accuracy of its facts, but such accuracy was nowhere
near good enough for release into the wild.
As such, and leaving aside for one moment Hume’s distinction between matters of fact and relations of ideas in An Enquiry Concerning
Human Understanding (to which truth as generally accepted reasoning
may be said to belong), this leaves us with a trade-protected page ranking system that depends on elements such as popularity as determined
by hyperlinks or aggregated searches, the exogenous links referred to
by Google’s engineers. Similarly, such aggregated data inform Google’s
autosuggest feature, helping searchers by beginning to complete their
search terms.
Carole Cadwalladr, writing in The Guardian in 2016, described her
experience of undertaking a thoroughly mundane, twenty-first-century
task: googling a search term. Entering the letters a-r-e j-e-w-s, Google
autocomplete offered a range of suggested questions: “are jews a race?”,
“are jews white?” “are jews christians?” and “are jews evil?”
36
Alongside a range of suggested answers to the questions (including that “Jews
are demonic souls from a different world”), Cadwaladr discovered that
the direct answer autocorrect suggestion to the word “women” was “are
women evil?”. In 2006, Google had been subject to a lawsuit by the
Anti-Defamation League because the first result for “jew” had been a
link to the anti-Semitic website, jewwatch.org, leading Google to emphasise that the page ranking system was automated. Over the next few
years, the processes involved in generating Google results continued to
attract attention from time to time, as when Tom Chatfield drew attention to the search engine’s disturbing trend of suggesting that women
shouldn’t work, shouldn’t vote, and shouldn’t even have rights.
37
In
general, Google had taken a lofty role with regard to its position, exonerating itself as a company with regard to its duties of overseeing its
algorithmic curators. In December 2016, however, Google was one of
a small handful of tech companies being blamed for the surprise outcome of the American election. Cadwalladr’s story was published on