An algorithm to detect outside influences on the media
Author: Sarah Aubort Source: Mediacom
EPFL researchers recently developed an
algorithm that maps out the media landscape and reveals biases and hidden
influences in the news industry.
News consumers may not be aware that the
way their local media outlet selects and presents news stories can be affected
by the media group that owns it. At a time of rampant disinformation, it is
just this sort of outside influence on the media that people should know about.
Researchers at EPFL’s Distributed
Information Systems Laboratory (LSIR) have come up with a way to make the news
industry more transparent. Their initiative, Media
Observatory,maps out the Swiss and international media landscape
through the topics that local media outlets choose to cover. It then uses those
choices to identify possible outside influences on the outlets. “The main
difficulty in this approach is the absence of an objective baseline: the simple
act of selecting stories is inherently biased. So we decided to compare
thousands of news sources and then map them out based on their similarities,”
says Jérémie Rappaz, an LSIR researcher and one of the main authors of the
study.
The researchers’ approach not only groups
news outlets geographically and by topic selection, but it also, crucially,
reveals the impact of the media groups that own them. Once mapped out, any
sudden change in the established editorial line of a given media outlet is
immediately apparent. “Most of these changes result from a change in ownership,
since media outlets tend to focus on topics favored by the group they belong
to,” says Rappaz. Yet news consumers don’t necessarily know which group their local
media outlet belongs to, if any. The EPFL researchers are seeking to fill this
gap by identifying links between media outlets through the content they publish
and by highlighting the extent of media concentration.
“Media groups in Switzerland and abroad are
increasingly pooling their resources in order to cut the cost of producing
stories. Unfortunately, media concentration of this type narrows the range of
views that consumers are exposed to. This is worrying, especially when the
views they get are biased,” says Rappaz. A striking example of this situation
was aired by John Oliver in a recent episode of Last Week Tonight in
the United States. It shows journalists on local TV channels owned by the
Sinclair Group all reading an identical
script commenting – ironically – on the lack of diversity in
the news.
Personalization tools
The Media Observatory has the backing of
the EPFL-based Initiative for Media Innovation (IMI) and has entered into
a partnership with the Swiss daily newspaper Le Temps. This support
will enable the LSIR team to bring its mapping project to the public next year
through an online platform. The website will model news production in
Switzerland and around the world while at the same time raising the public’s
awareness of the dangers of disinformation. “We are pleased to be a part of the
LSIR project,” says Gaël Hurlimann, co-editor in chief and head of digital
at Le Temps. “We really want to understand, and help the public
understand, what media concentration means for all of us. It’s important to be
transparent about the factors influencing news production: that’s the only way
we can build trust between ourselves – the media – and our consumers. And who
knows, maybe the study’s findings will convince us to make some changes in how
we work.”
The web-based platform will use
transparent, open-source technology. It will be built on personalization
algorithms, similar to the ones used by websites like Netflix and Amazon Prime,
which suggest videos based on people’s viewing history. “We applied this
concept to media coverage data and were very surprised by what we could do with
it,” says Rappaz. The researchers fed the algorithm around 500 million articles
published by 8,000 different sources over the past three years. The algorithm
assessed how individual news stories were handled by different media outlets,
grouping the outlets on the basis of their similarities and revealing links
between them and influences on them.
The LSIR team hopes that its research and
web platform will encourage people to take a more critical approach to the news
they consume, and help journalists investigating how stories are covered. For
more information on the project: https://www.mediaobservatory.com/
References
A Dynamic Embedding Model of the Media
Landscape, Jeremie Rappaz, Dylan Bourgeois, Karl Aberer, The Web Conference
2019 (WWW '19) https://www.mediaobservatory.com/assets/papers/www19.pdf
Nessun commento:
Posta un commento