Hey Google, what was the Holocaust about? Auditing how search engine algorithms structure memories about mass atrocities

Makhortykh, Mykola; Urman, Aleksandra; Ulloa, Roberto (2022). Hey Google, what was the Holocaust about? Auditing how search engine algorithms structure memories about mass atrocities (Unpublished). In: Communicating Memory Matters: Next Steps in the Study of Media Remembering and Communicative Commemoration. Salzburg. June 30-July 1 2022.

Full text not available from this repository.

In today’s algorithm-driven media ecologies, search engines such as Google or Baidu serve as key modulators of public knowledge about the present and the past. Using big data and machine learning techniques, search algorithms allow users to tackle the information overload by filtering, ranking and personalizing information in response to their queries. However, by doing so, search engines become hegemonic gate-keepers determining what information is consumed by the citizens (Noble 2018).
With a few exceptions (e.g. Zavadski & Toepfl 2019), there is little research on how search engines structure historical knowledge. Yet, this question is of paramount importance considering the ongoing “algorithmic” (Esposito 2017) memory turn, under which collective perceptions of the past are increasingly shaped by search algorithms. By deciding what kind of sources to promote and whether to give visibility to revisionist/negationist narratives, search engines create hierarchies of historical knowledge. Similarly, the prioritization of specific pieces of (audio)visual content has significant implications for the cultural lexicon of symbols used to represent the past (Stier 2015).
In our paper, we introduce a novel approach for large-scale comparative analysis of algorithmic structuring of historical knowledge. Building upon earlier studies on algorithmic auditing (e.g. Hannák et al. 2017; Kulshrestha et al. 2017; Puschmann 2019), we develop a distributed cloud-based research infrastructure for emulating and recording the browsing behaviour of multiple autonomous agents. By enabling control over specific variables (e.g. browser type), our infrastructure allows identifying the effect of different factors on how memory-related content is structured and prioritized.
As a case study, we looked at how memory of three cases of mass attrocities committed in the 20th century - the Armenian Genocide, the Holodomor and the Holocaust - is structured in six major search engines, i.e Google, Yandex, Bing, Yahoo, DuckDuckGo and Baidu. Using queries in English, Russian and Chinese (the selection of languages is attributed to the affiliation of corporations behind the search engines) and a large number of autonomous agents (n=200), we collected for each agent the first 50 search results together with the first 50 images and the 50 videos suggested by a specific search engine in relation to the three events. While doing so, we took several measures to isolate the effect of search personalization; specifically, we used dynamic cloud-based IPs for each agent, synchronised their activity to isolate the factor of time and compared it across two browsers (Chrome and Firefox).
Using the collected data, we then discuss the following questions: In which ways do the hierarchies of historical sources differ between the search engines and how are these hierarchies affected by the language of the query and/or external variables (e.g. browser type)? How significant is the effect of randomization on the ranking of atrocities-related content within a specific search engine and what are the consequences for knowledge structuring? How different is prioritization of (audio)visual representations of the mass atrocities between the search engines and what aspects of the events are promoted by these algorithmically “iconized” images?
References
Esposito, E. (2017). Algorithmic memory and the right to be forgotten on the web. Big Data & Society, 4(1), 2053951717703996.
Hannák, A., Sapieżyński, P., Khaki, A. M., Lazer, D., Mislove, A., & Wilson, C. (2017). Measuring Personalization of Web Search. ArXiv:1706.05011 [Cs]. http://arxiv.org/abs/1706.05011
Kulshrestha, J., Eslami, M., Messias, J., Zafar, M. B., Ghosh, S., Gummadi, K. P., & Karahalios, K. (2017). Quantifying Search Bias: Investigating Sources of Bias for Political Searches in Social Media. Proceedings of the 2017 ACM Conference on Computer Supported Cooperative Work and Social Computing, 417–432. https://doi.org/10.1145/2998181.2998321
Noble, S. U. (2018). Algorithms of oppression: How search engines reinforce racism. NYU Press.
Puschmann, C. (2019). Beyond the bubble: Assessing the diversity of political search results. Digital Journalism, 7(6), 824-843.
Stier, O. B. (2015). Holocaust icons: Symbolizing the Shoah in history and memory. Rutgers University Press.
Zavadski, A., & Toepfl, F. (2019). Querying the Internet as a mnemonic practice: how search engines mediate four types of past events in Russia. Media, Culture & Society, 41(1), 21-37.

Item Type:	Conference or Workshop Item (Paper)
Division/Institute:	03 Faculty of Business, Economics and Social Sciences > Social Sciences > Institute of Communication and Media Studies (ICMB)
UniBE Contributor:	Makhortykh, Mykola, Urman, Aleksandra
Subjects:	000 Computer science, knowledge & systems 300 Social sciences, sociology & anthropology 900 History 900 History > 940 History of Europe
Language:	English
Submitter:	Mykola Makhortykh
Date Deposited:	17 Aug 2022 08:23
Last Modified:	05 Dec 2022 16:22
Additional Information:	Holocaust, algorithm audit, digital memory, Google, bias, history
URI:	https://boris.unibe.ch/id/eprint/171974