106 Turing’s Test
• Potential: This especially relates to the ability of software to generate significant amounts of content and on demand.
• Limitations: Algorithms cannot query the data that are provided,
and at present the writing capabilities of software are limited although likely to improve considerably over the next decade.
24
In addition, the Tow Center report considers the implications for
journalists – especially with regard to employment, what Montal and
Reich and other commentators have outlined as the occupational risk –
but also the possible benefits that could come through the automation
of routine tasks. Interestingly, while Montal and Reich considered algorithmic transparency as an essential moral question for organisations,
Graefe’s own research did not indicate whether there was a significant
demand for such transparency on the part of consumers,
25
although that
research was conducted prior to the 2016 Presidential election and the
subsequent raising of public awareness with regard to fake news. It is possible, however, that although consumers are – as Graefe observes – less
likely to enjoy reading automatically generated content, they may still believe it more credible if it is labelled as such. Transparency does become
crucial for organisations when errors emerge, but the bigger questions –
largely unanswered by Graefe – revolve around the role of journalism as
a watchdog of the state if it becomes increasingly automated.
With regard to the drivers and conditions for the increase of automation within the industry, Dörr in an article published in the same year
as Graefe’s report highlights the reasons for publishers’ keen interest:
the struggle for profitability, market share, and readers.
26
As with so
many other areas, the primary driver for increased automation is to cut
costs, but the relations between journalism and technology have always
been close and, as we shall see, I believe that automation offers potential
advantages to journalists as well as news organisations. In any case, the
limitations of algorithms should not be neglected, especially with regard
to their requirements for highly structured data to be effective. We have
already seen that a search over a period of a year for the phrase “This
story was generated by Automated Insights” returned just under 90,000
stories: in the same period, there were more than 29 million stories and
articles referencing Trump, none of which appeared to have been written
by an algorithm. In the field of journalism, at least, Donald Trump has
kept his promise to bring back jobs to American workers.
Machine Learning and Language
A common refrain as Trump’s presidency began has been to comment
on how irregular the White House’s operations were under his tenure.
Francine Prose, writing for The Guardian, criticised his activities under
the headline “Nothing about the Trump presidency is normal”,
27
while
Walter Shapiro observed wryly: “Never Normal Is the New Normal at
• Potential: This especially relates to the ability of software to generate significant amounts of content and on demand.
• Limitations: Algorithms cannot query the data that are provided,
and at present the writing capabilities of software are limited although likely to improve considerably over the next decade.
24
In addition, the Tow Center report considers the implications for
journalists – especially with regard to employment, what Montal and
Reich and other commentators have outlined as the occupational risk –
but also the possible benefits that could come through the automation
of routine tasks. Interestingly, while Montal and Reich considered algorithmic transparency as an essential moral question for organisations,
Graefe’s own research did not indicate whether there was a significant
demand for such transparency on the part of consumers,
25
although that
research was conducted prior to the 2016 Presidential election and the
subsequent raising of public awareness with regard to fake news. It is possible, however, that although consumers are – as Graefe observes – less
likely to enjoy reading automatically generated content, they may still believe it more credible if it is labelled as such. Transparency does become
crucial for organisations when errors emerge, but the bigger questions –
largely unanswered by Graefe – revolve around the role of journalism as
a watchdog of the state if it becomes increasingly automated.
With regard to the drivers and conditions for the increase of automation within the industry, Dörr in an article published in the same year
as Graefe’s report highlights the reasons for publishers’ keen interest:
the struggle for profitability, market share, and readers.
26
As with so
many other areas, the primary driver for increased automation is to cut
costs, but the relations between journalism and technology have always
been close and, as we shall see, I believe that automation offers potential
advantages to journalists as well as news organisations. In any case, the
limitations of algorithms should not be neglected, especially with regard
to their requirements for highly structured data to be effective. We have
already seen that a search over a period of a year for the phrase “This
story was generated by Automated Insights” returned just under 90,000
stories: in the same period, there were more than 29 million stories and
articles referencing Trump, none of which appeared to have been written
by an algorithm. In the field of journalism, at least, Donald Trump has
kept his promise to bring back jobs to American workers.
Machine Learning and Language
A common refrain as Trump’s presidency began has been to comment
on how irregular the White House’s operations were under his tenure.
Francine Prose, writing for The Guardian, criticised his activities under
the headline “Nothing about the Trump presidency is normal”,
27
while
Walter Shapiro observed wryly: “Never Normal Is the New Normal at