-
Notifications
You must be signed in to change notification settings - Fork 0
/
index.html
372 lines (295 loc) · 18.3 KB
/
index.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
<!DOCTYPE HTML>
<!--
Phantom by HTML5 UP
html5up.net | @ajlkn
Free for personal and commercial use under the CCA 3.0 license (html5up.net/license)
-->
<html>
<head>
<!-- Global site tag (gtag.js) - Google Analytics -->
<script async src="https://www.googletagmanager.com/gtag/js?id=UA-116401611-3"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'UA-116401611-3');
</script>
<title>Media Observatory - Quantifying Media Bias</title>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1, user-scalable=no" />
<link rel="stylesheet" href="assets/css/main.css" />
<noscript><link rel="stylesheet" href="assets/css/noscript.css" /></noscript>
<meta name="twitter:card" content="summary_large_image">
<meta name="twitter:creator" content="@JRappaz">
<meta name="twitter:title" content="Quantifying Media Bias: The Influence of Ownership on Media Coverage">
<meta name="twitter:description" content="We produced a dynamic map of the media landscape over the last 3 years.">
<!--<meta name="twitter:image" content="https://www.mediaobservatory.com/images/banner_selbias.png">-->
<meta name="twitter:image" content="https://www.mediaobservatory.com/images/concentration.jpg">
<meta property="og:image" content="https://www.mediaobservatory.com/images/concentration.jpg">
<meta property="og:image:type" content="image/jpeg">
<meta property="og:image:width" content="2724">
<meta property="og:image:height" content="1089">
<style type="text/css">
.logo {width: 120px;}
#top-insights {
font-weight: 600;
}
#main {margin-top: 40px;}
.pad-left{padding-left: 70px;}
.highred{color: #f2849e}
h4 { margin-bottom: 0.5em;}
</style>
</head>
<body class="is-preload">
<!-- Wrapper -->
<div id="wrapper">
<!-- Main -->
<div id="main">
<div class="inner">
<header>
<img src="images/epfl.png" style="width: 150px">
<h1>Mapping the News Landscape</h1>
<h2>The Influence of Ownership on Media Coverage</h2>
<!--<h3>by <span style='color:#aaa'><a href="https://jrappaz.com" target="_blank">Jérémie Rappaz</a>,
<a href="https://dtsbourg.me" target="_blank">Dylan Bourgeois</a>,
<a href="https://lsir.epfl.ch" target="_blank">Karl Aberer</a></span> @ <span style='color:#aaa'>WebConf 2019 </span>
</h3>-->
<!-- <p>Accepted for an oral presentation at WebConf 2019 - Web & Society track</p> -->
</header>
<br /><br />
<figure id="fig1">
<img src="images/banner_selbias.png"/>
<br /><br />
<figcaption>
<div class="fig-title">Figure 1: A comprehensive map of the media landscape.</div>
</figcaption>
</figure>
<br />
<!--<b>Figure</b>
<i>Visualizing the media landscape: sources with similar coverage are closer together.
Clusters are annotated when as belonging to larger groups that the method is able to
capture through coverage similarities.</i>
<br /><br />
<span class="image right"><img src="images/news.gif" alt="">Animation of a news source (krcrtv.com) which was acquired in December 2017</span>
-->
<section>
<h3 class="underlined">OUR MISSION</h3>
<p>
Media bias is hard to quantify since it is generally subjective and heavily depends on human perceptions.
We therefore developped a fully automated method that maps the digital news landscape, based on historical publishing data.
In short,
</p>
<div id="top-insights">
<ul>
<li> We propose a model of the media landscape that automatically maps news sources based on their selection of subjects.</li>
<li> By tracking sources' evolution over time, we identify driving forces, from the influence of ownership to large-scale content diffusion patterns.</li>
<li> We produce a dynamic map of the media landscape over the last 3 years, based on coverage from 500M articles shared by around 8K sources.</li>
</ul>
</div>
We plan to release an open and interactive map based on these insights, which can be used by anyone, from viewers to journalists, to dig into patterns that influence the media ecosystem. <a href="#contact">Subscribe to our mailing list to stay tuned!</a>
</section>
<br /><br />
<section>
<h3 class="underlined">THE PROBLEM</h3>
<p>The digitalization and globalization of our news channels have had a profound impact on worldwide information diffusion. Nowadays, information
is processed in mass by large organizational structures, before being propagated to local channels.
This process often obfuscates the origin of a news story, blurs the accountability chain to the author of its coverage,
and reduces the traceability to its inception.
<br /><br />
Large broadcasting groups have recently been suspected of massive consolidation of resources, pooling resources to reduce the cost of producing stories.
Unfortunately, this reduces the diversity of coverage that reaches the viewer, with multiple local channels sharing the same
selection of news, pruned by actors often unbeknownst to them.
The lack of diversity and accountability can be worrisome, particularly when this content is ideologically charged, as has recently <a href="https://www.youtube.com/watch?v=GvtNyOzGogc">been exposed by John Oliver</a> in his show <i>Last Week Tonight</i> (HBO). This surreal footage shows journalists all reading the same scripted text, provided by Sinclair Broadcasting Group, a large conglomerate of local sources, about ... the lack of diversity in the news.
<iframe width="728" height="410" src="https://www.youtube.com/embed/hWLjYJ4BzvI" frameborder="0" allow="accelerometer; autoplay; encrypted-media; gyroscope; picture-in-picture" allowfullscreen></iframe>
<!--This phenomenon has an impact on what is called the <a href="https://www.mediamonitor.nl/english/the-concept-of-pluralism-media-diversity/">external pluralism</a> assumption: theoretically, a large number of news channels should be available to ensure a variety of viewpoints, but the suggested rising concentration of media ownership may void this assumption. This observation motivates the study of the impact of ownership on the global media landscape and its influence on the coverage the actual viewer receives.
<br /><br />-->
In our work, we proposed an algorithm to automate the tracking and the monitoring of broadcasted news content.
One of the main difficulties in the development of such a method is the absence of ground truth:
any selection of events is partial by nature and is only comparable to the selection made by other channels.
Therefore, our method compares the coverage of events <i>across</i> news entities. This produces a similarity map, where distances
between channels represent the similarity of their coverage. An example map is shown in <a href="#fig1">Figure 1</a>.
<br /><br />
The map reveals a structure that is directly derived from coverage. This means we can identify common patterns in sources' coverage after the fact, i.e. without any assumptions on
what they may be. In other words, we do not bias the search for biases.
Through analysis, we uncovered topical factors such as sources' geographic relationships - local sources cover similar events - or thematic regards.
Interestingly, we also reveal the influence of higher-order broadcast structures, evidence for the influence of large corporate actors in local news.
Isolating these patterns is far from obvious, particularly given that a source's branding does not always reveal its affiliation: large broadcast entities
like Sinclair own franchised outlets, often household names like Fox or ABC. Our method provides transparency to the viewer, going beyond surface branding to
reinstate accountability towards the true origin of the broadcast content.
<!-- However, many of these outlets are actually owned by larger corporate structures, some of which, <a href="https://www.youtube.com/watch?v=GvtNyOzGogc" target="_blank">most infamously Sinclair Broadcast Group</a>, have been accused of imposing ideologically charged content on these downstream channels, unbeknownst to the casual viewer. -->
<br /><br />
Technically speaking, our method borrows from the field of personalization with a set of techniques commonly used in recommender systems,
such as those used on Amazon or Netflix. The intuition is that we can model a source's selection of covered news as a preference problem:
the editor <i>selects</i> or chooses the events that are most relevant to their audience.
As such, the algorithm tries to predict the likelihood of a channel covering a given event. To do so, it observes
which other sources have selected it, trying to mimic the behavior of sources similar to that which it is trying to model.
In a sense, the algorithm learns a fingerprint of each sources' coverage patterns so it can match it to other similar ones.
These coverage profiles can then be compared to uncover similarities in the choice of their reporting,
usually indicating the presence of a common bias. In sum, sources appearing close together on the map are likely to have a very similar selection of events.
</p>
</section>
<section >
<h3 class="underlined">An EVOLVING MAP</h3>
<figure id="fig2">
<img src="images/evolution.png"/>
<br /><br />
<figcaption>
<div class="fig-title">Figure 2: An evolving map of the media landscape.</div>
</figcaption>
</figure>
<br />
<p>
We have shown how we can generate a comprehensive map of the news ecosystem from a selection of reported events. However, this first method only
provides a static view of an inherently dynamic system. As the media landscape is always evolving, we need a more complete model which
takes into account the dynamics of the news ecosystem. We extend the ideas that provide us with a static snapshot to produce an evolving map of source
similarity. It highlights
important changes that occur when channels modify their news coverage habits. The external or internal forces driving this change can also
be isolated to better understand influences on the media ecosystem.
<br /><br />
For example, we demonstrate the method's potential for news monitoring applications and investigative journalism by shedding
light on important changes in programming induced by mergers and acquisitions, policy changes, or network-wide content
diffusion. An example of the phenomenon is shown <a href="#fig2">in Figure 2</a>: news sources tend to agglomerate into
cohesive clusters after their acquisitions by three well-known news companies, clearly showing the consolidation of resources
taking place in the background and the uniformization of information channels.
</p>
</section>
<section>
<h3 class="underlined">DETECTING HIGHLY INFLUENCIAL BROADCAST GROUPS</h3>
<figure id="fig3">
<img src="images/concentration.jpg"/>
<br /><br />
<figcaption>
<div class="fig-title">Figure 3: Clustering highly influential sources, tracking alignment with the sources they attract.</div>
</figcaption>
</figure>
<br />
<p>
One of the most common observable patterns is the convergence of a group of sources towards a specific position: as the similarity of
their selection of news increases, the model will highlight their increased similarity. We propose an automated method to detect this
phenomenon by introducing the notion of <i>attractors</i>: the algorithm identifies regions of the map toward which many sources converge.
Regions with a high potential of attraction are shown in blue in <a href="#fig3">Figure 3</a>.
<br /><br />
Once identified, the attracted news sources, which are converging towards an attractor, can be monitored. For example, the acquisition
campaign performed by Sinclair in September 2017 can be visualized on the right-hand side of <a href="#fig3">Figure 3</a>. The distance from
four sources to the attractor is abruptly reduced, suggesting an alignment of content post-acquisition.
</p>
</section>
<section>
<h3 class="underlined">PUBLICATIONS</h3>
<ul>
<li>
<b><a href="#" class="icon fa-file" style="margin-right:1em;"><span class="label">Paper</span></a>
A Dynamic Embedding Model of the Media Landscape </b>   //   <span class="highred">WWW'19</span>
<li class="pad-left">
<a href="assets/papers/www19.pdf">Paper</a>
<!--<input type="submit" value="paper" class="small">
<input type="submit" value="code" class="primary small">
<input type="submit" value="bib file" class="primary small"></li>-->
</li>
<li>
<b><a href="#" class="icon fa-file" style="margin-right:1em;"><span class="label">Paper</span></a>
Selection Bias in News Coverage: Learning it, Fighting it </b>   //   <span class="highred">WWW'18</span>
<li class="pad-left">
<a href="assets/papers/www18.pdf">Paper</a>
<!--<input type="submit" value="paper" class="small">
<input type="submit" value="code" class="primary small">
<input type="submit" value="bib file" class="primary small">-->
</li>
</li>
</ul>
</section>
<!-- Image -->
<section>
<h3 class="underlined" id="contact">CONTACTS</h3>
Don't hesitate to reach out to <b><a href="mailto:[email protected]">[email protected]</a></b> for more information or resources regarding this work.
To follow our research, and stay informed of new work, for example, the upcoming interactive media map, you can register for updates here:
<br /><br /><br />
<form name="submit-to-google-sheet">
<div class="container" style="background-color:white; float: left; width: 50%; margin-left: 15%;">
<input type="email" placeholder="Email address" name="email" required style="height: 2.8em;border-radius: 4px 0 0 4px; padding-left: 30px;">
</div>
<div class="container" style="float: left;">
<input type="submit" value="Subscribe" style="border-radius: 0 4px 4px 0;">
</div>
</form>
<br />
<h5>Members of the initiative</h5>
<div class="box alt">
<div class="row gtr-uniform">
<div class="col-2"><span class="image fit"><img src="images/jay.jpg" alt="" /></span></div>
<div class="col-10">
<h4><a href="https://jrappaz.com">Jérémie Rappaz</a></h4>
<p>
Ph.D. student at EPFL. His research interests focus on predictive methods that involve massive amounts of people.
</p>
</div>
</div>
</div>
<div class="box alt">
<div class="row gtr-uniform">
<div class="col-2"><span class="image fit"><img src="images/dylan.png" alt="" /></span></div>
<div class="col-10">
<h4><a href="https://dtsbourg.me">Dylan Bourgeois</a></h4>
<p>
Masters Student at EPFL, currently working towards his masters’ thesis at Stanford University.
He focuses on the application and implications of Machine Learning in different disciplines,
with work in news bias characterisation, robotics or high-energy physics at CERN.
</p>
</div>
</div>
</div>
<div class="box alt">
<div class="row gtr-uniform">
<div class="col-2"><span class="image fit"><img src="images/karl.jpg" alt="" /></span></div>
<div class="col-10">
<h4><a href="https://lsir.epfl.ch">Karl Aberer</a></h4>
<p>
Karl is a full professor at EPFL and head of the Distributed Information Systems Laboratory.
</p>
</div>
</div>
</div>
<!--
<br /><br /><br />
<br /><br /><br /> -->
</div>
</div>
<!-- Footer -->
<footer id="footer">
<div class="inner">
<!-- <section>
<h2>Get in touch</h2>
<form method="post" action="#">
<div class="fields">
<div class="field half">
<input type="text" name="name" id="name" placeholder="Name" />
</div>
<div class="field half">
<input type="email" name="email" id="email" placeholder="Email" />
</div>
<div class="field">
<textarea name="message" id="message" placeholder="Message"></textarea>
</div>
</div>
<ul class="actions">
<li><input type="submit" value="Send" class="primary" /></li>
</ul>
</form>
</section> -->
<!-- <section>
</section> -->
<ul class="copyright">
<li>© LSIR. All rights reserved</li><li>Template: <a href="https://html5up.net">HTML5 UP</a></li>
</ul>
</div>
</footer>
</div>
<!-- Scripts -->
<script src="assets/js/jquery.min.js"></script>
<script src="assets/js/browser.min.js"></script>
<script src="assets/js/breakpoints.min.js"></script>
<script src="assets/js/util.js"></script>
<script src="assets/js/main.js"></script>
<script src="assets/js/gform.js"></script>
</body>
</html>