This story draft by @escholar has not been reviewed by an editor, YET.

This is what a pandemic looks like: Visual framing of COVID-19 on search engines: Findings

EScholar: Electronic Academic Papers for Scholars HackerNoon profile picture

This paper is available on arxiv under CC 4.0 license.


(1) Mykola Makhortykh, Institute of Communication and Media Studies, University of Bern;

(2) Aleksandra Urman, Social Computing Group, University of Zurich;

(3) Roberto Ulloa, GESIS – Leibniz-Institute for the Social Sciences.

Table of Links

Author Note


Theoretical background



Conclusion and References


The first finding is that the number of actual frames in the search results is small compared with the usual number of frames used by legacy media for framing the pandemics (Kee, Faridah, and Normah 2010; Pan and Meng 2016). As shown in Table 1, the ratio of frame strength is low for the majority of the search engines. In the case of search queries for Bing in English, we do not find any generic news frames at all, whereas in other cases, the number of images which can be treated as frames is extremely low. Instead of news frames in the usual sense of the word, many images prioritized by the search engines, in particular for English language, are schematic depictions of coronavirus particles. Examples of such images are shown in Figure 1 below. The major difference between such images is the color of the coronavirus particles and the level of the detail.

Figure 1. Schematic depictions of coronavirus search results. The images are taken from Unsplash and attributed to the Public Health Image Library from the Centers for Disease Control and Prevention.

The common feature of these non-frame images is their lack of interpretability that limits their use as a means of interpreting what coronavirus is. By themselves, the schematic depictions of coronavirus do not provide the user with information that is helpful for their understanding of the issue at hand. Not only are such images very different from generic news frames, but also their prevalence seems to fulfill only the most basic functionality of the search engine, i.e., finding the most specific images that illustrate the topic (e.g., virus particles in the case of COVID-19, which is not necessarily optimal for keeping users informed.

Table 1. Frame strength per engine-language combination (mean values)

The second finding is the substantially higher strength of the human-interest frame compared to the other four frames. The most common appearance of this frame involves giving a human example or “human face” to the pandemic. Often, this representation of COVID-19 is supplemented by the emphasis on the pandemic’s effect on individuals and societies. Usually, such frames picture masked medical personnel and/or the patients. In some cases (in particular, the queries in Chinese), it also includes schematic instructions explaining how COVID-19 is distributed in order to inform the public. An interesting case of the human-interest frame is observed for queries in Russian from Western search engines. There, the human-interest frame is dominated by visuals generating the feeling of empathy by showing images of felines, often with tears in their eyes or sitting on the hands of medics. The high visibility of felines can be related to search queries concerning the possibility of cats getting infected with COVID-19 which for some reason occurred particularly often in Russian.

Despite the high visibility of the responsibility frame in the case of legacy media coverage of other pandemics (Kee, Faridah, and Normah 2010), this frame occurs rarely in the case of COVID-19 framing on search engines. The other three types of frames are more underrepresented and appear just in a few engine-language combinations. This specific distribution of frames can be related to the differences in the framing process on search engines and legacy media, but also to differences in the framing of COVID-19 and earlier pandemics, so further research is required to clarify them.

The third and the final finding points to substantial differences in the framing of COVID-19 on various search engines and in various languages. To examine these differences, we aggregated data on the average frame strength for each particular engine (Table 2) and each language in which we conducted searches (Table 3).

The cross-language comparison indicates that two groups of engines can be distinguished based on their framing of COVID-19. The first of them consists of Bing, Yahoo, and DuckDuckGo, where the number of frames in the search results is low and the results themselves are mostly represented by the schematic images of coronavirus particles (see Figure 1). The second group composed of Google and Yandex has more actual frames among the search results, in particular the ones related to responsibility and human interest. Furthermore, only engines from the second group include the morality frame. These differences can be attributed to several factors which are primarily related to the implementation of the ranking and filtering algorithms by different search engines. In the case of image search, these algorithms have different potential for recognizing the images and then contextualizing them to respond to the text queries. Some of these algorithms share substantial similarities (e.g., in the case of Bing, Yahoo and DuckDuckGo (Chris 2020)), whereas others (e.g., Google and Yandex) develop their unique solutions. Another explanation can be related to the differences in the type and the size of the audience utilizing the respective engines, which influences the quality of data used to filter and rank content for the respective queries.

Table 3. Cross-language difference in frame strength (mean values)

Similarly, we observe a number of differences among the languages in which the queries were conducted. Specifically, almost all generic frames tend to be stronger in Russian and Chinese compared with English. The strength of the human-interest frame is particularly more pronounced with the queries in the former two languages returning more images of human actors in relation to COVID-19 as contrasted with more schematic depictions of the virus in English.

As in the case of cross-engine differences, the reasons for discrepancies among languages can be explained by several reasons. The stronger presence of the human-interest frame for Chinese queries can be attributed to China taking the strongest hit from the pandemic at the time of data collection (end of February 2020). Hence, for China, the human toll of COVID-19 was more visible compared with Western countries, where the number of infected was considerably lower. While in Russia, where the number of COVID-19 cases was also low, there seemed to be a disproportionate amount of interest in the consequences of the pandemic for house pets (in particular, cats), that together with the recognition of the emergency in areas neighboring China could contribute to the rise of awareness about the threat of COVID-19.

. . . comments & more!

About Author

EScholar: Electronic Academic Papers for Scholars HackerNoon profile picture
EScholar: Electronic Academic Papers for Scholars@escholar
We publish the best academic work (that's too often lost to peer reviews & the TA's desk) to the global tech community


Around The Web...