The visual patterns of a phenomenon

Google Image Search as a useful ally to visually explore the migrant crisis controversy

Whenever we want to find out more about major events that take place around us every day, from wars to natural calamities, it’s fair to say that we turn to news organizations for up-to-date informations and developments. A great share of people still rely heavily if not entirely on them to get informed on what is going on around the world. Moreover, the things that we remember the most of a given issue are often the images that we have seen on TV or on the Internet. Images with the power to shape our perception of a phenomenon and that often differ from one person to the other, given the huge amount of pictures the media industry provides us everyday. It’s important to keep in mind that every photograph is the result of an editorial decision and the amount of images related to an issue reflect the multitude of opinion on the same issue. As Susan Sontag wrote (Regarding the Pain of Others, 2003):

“to photograph is to frame, and to frame is to exclude”

Images have the potential to draw attention to a particular phenomenon and even influence the debate that surrounds it.

This is especially true for complex issues such as the European migrant crisis. The discussion related to it has been deeply influenced by images, for example by the picture of Aylan Kurdi, a Syrian child found dead on the coast of Bodrum, Turkey (a very interesting analysis has been conducted by researchers at the University of Sheffield’s Visual Social Media Lab about the impact this single image had on the debate on immigration, you can find it here) and this debate differs from country to country throughout the European Union. We asked ourselves: Which photographs related to this crisis circulated the most in each EU country? Which of them shaped the most the public opinion? Do different organizations use the same photographs? Can we unveil precise editorial patterns behind news organizations’ narratives? Using Google Image Search, we tried to investigate and understand how the biggest news organizations in Europe depicted the European migrant crisis through pictures, as well as how the issue was perceived in each EU Member State.

Actors definition

In order to select the most read online news websites, hence the ones that potentially have the most influence on people, we used Alexa’s Top Sites Rank, picking the top two sites that produced the news as their primary activity for each EU country. This resulted in 56 actors that would later become the targets of the google advanced search thanks to the query operator “site:”.

(more information about Google advanced search query operators)

Queries definition

When referring to the European migrant crisis, the four most queried terms on Google were: refugee, migrant, refugee crisis and migrant crisis. Since almost every country in Europe speaks its own language though, we translated this four terms in 23 languages. Beginning from the wikipedia article “European migrant crisis” in its English version, we traced back the terms in the other available languages. For the languages that were missing, we used the Wayback Machine to jump back in time on a news organization website, in order to find a series of articles related to the migration issue and therefore discover (thanks to Google Translator) the right words used in that language. It’s important to note that not every country made extensive use of the four terms, so for each Member State we evaluated if the queries were all necessaries or if any of them would have just created unwanted noise.

Image scraping

Using Firefox’s or Google Chrome’s private navigation (Incognito Mode) we queried Google Images in the following way:

The query operator “daterange:” refers to our need to only focus our research in 2015, so that Google returned for each news’ website the images its algorithm found to be the most relevant when the debate over the issue was at its peak. For each of the 56 actors, we downloaded the first 80 images with DownThemAll! and we later kept only the first 50 keeping only photographs (no maps, graphs or hybrid images).

In order to get the images’ links to their related articles (they could always become useful for further analysis, for example language analysis) we extracted the HTML code of each Google search and finally separated the links using TextWrangler and some RegExps. At the end of this process we were left with 2800 photographs, 50 for each news organization so 100 for each country. A quick way to look at them all together is with ImageSorter, which automatically places a stack of pictures close together by hue.

Image categorization

With this corpus in our hands, we proceeded with two parallel tasks: the first one to know which images were used by more than one actor and if there were, how many times they were used; the second one to determine what the most depicted subjects were in different news’ sites. The former was accomplished with Photosweeper X, a software normally intended to get rid of duplicate photos but that this time helped us single out 46 groups composed by at least 3 identical (or almost) photos. For what concerns the latter, we went through the corpus repeatedly one image at a time, giving each of them tags based on what they were portraying. We defined 21 tags that recurred several times in our corpus: children, volunteers, rescuer, outdoor encampment, banners and flags, garbage, forms of transport, emergency shelter, blood, sea, law enforcement, barrier, road, raft, rails, weapons, field, emergency equipment, journalist.

During the process we noticed that for some tag, namely law enforcement, a further subdivision was needed. As a matter of fact, the police was portrayed quite differently: from the kindest gesture to the most brutal repression. This is why we divided furthermore the law enforcement tag in three different concepts: law enforcement helping people, law enforcement monitoring and law enforcement fighting.

So, to give a brief summary:

Findings

Taking a closer look at the images, we can say that there are several groups of images that recur in different actors. Events such as the death of Aylan Kurdi (present at least once in 16 of the media sites we considered), the Syrian family surrounded by police on train tracks in Bicske (11 sites) and Hungarian journalist Petra Laszlo tripping a Syrian man and his son (11 sites) were the most recurrent events in our corpus.

There were a couple of things that caught our eyes though: first of all, the same event was not always portrayed with the same or similar images, but rather with angles or moments in time quite different. For example, when reporting the tragic event of Aylan Kurdi, the majority of media decided to represent either the moment his body was found by turkish police or the following minutes when one of the officers picked up the body. The position of the child’s body and the role of the police are radically divergent from one picture to the other, and our reaction to them varies as well, ranging from compassion to even anger.

As for the happening in Bicske, the perspective from which the pictures were taken differs highly and the sense of oppression seems to increase with the number of policemen in the picture.

Sometimes even if the original photo is the same, in the final article the image appears distorted or modified. The most blatant example is the use of a picture taken by italian photographer Massimo Sestini off the coasts of Lybia. The photo was actually shot in 2014, but it has been used since then to portray the mass arrivals of migrants from Africa, sometimes even stretching the picture to make the boat bigger.

Another interesting finding is that different pictures, even if they aren’t related to the same event, depict the same situation or present the same elements.

We observed that some of these elements had more relevance in some Countries than in others: every media tends to give more importance to the issues that its audience feels closer. Even if the representation of boats and dinghies was a relevant factor throughout almost all news’ sites, in Countries which had to deal with migrants and refugees arriving by sea (like Italy, Greece, Portugal and Cyprus) they were also the ones that represented boats and dinghies the most. To a minor extent, Countries that find themselves on the so-called “Balkan route” to Germany (Austria, Bulgaria, Greece, Czech Republic, Romania, Slovakia and Hungary) were more likely to show images of emergency outdoor encampment. A peculiar case was Denmark: its news organizations were depicting the migrants’ journey on the road by foot far more than any other news site in our corpus.

Other elements such as the presence of police forces were reported by a lot of media aside from any geographical factor. Also children were a recurring subject in the corpus with outstanding frequency: their purity and vulnerability are without doubt a very powerful method to lure the user into the news’ website.

The act of trespassing a national border underneath and over barbed wire or the agitated rush to board trains directed to northern Europe are other examples of situations represented multiple times.

Overall we can say that visually news organizations tend to favor images that always depict the same set of situations or elements, the ones that can produce the strongest emotions to their audience. Even when reporting single events, photos are being cropped and modified in order to amplify as much as possible the feeling they provoke. This often results in a representation of the phenomenon that is partial if not misleading.

This first visual investigation is just a foretaste of a much bigger world of explorations that could be done beginning from this corpus, for example taking an in-depth look at the language used together with these images or maybe comparing them with images associated to other migration crises (in a different part of the world or in the past).

Feasibility for other circumstances

The power and flexibility of the Google Image Search engine, combined with the appropriate allies, allows us to easily gather huge amounts of images, filtered by specific topics, domains, sites or periods of time. This protocol is helpful anytime we want to grasp how a particular issue is perceived by the general public (we can say Google’s search engine is a good approximation, since it retrieves what people ask for) in terms of images and photographs. Keep in mind that the results will be somewhat relevant only for issues where images play a key role in the developing of the debate. For what concerns the results/effort ratio, the protocol doesn’t take too much time to finish and if the image’s corpus doesn’t get too big, the categorization can be easly completed by hand. Otherwise services like Amazon Mechanical Turk or Crowdflower can become handy with only a small investment.