Do not blame it on the algorithm: an empirical assessment of multiple recommender systems and their impact on content diversity

Abstract

In the debate about filter bubbles caused by algorithmic news recommendation, the conceptualization of the two core concepts in this debate, diversity and algorithms, has received little attention in social scientific research. This paper examines the effect of multiple recommender systems on different diversity dimensions. To this end, it maps different values that diversity can serve, and a respective set of criteria that characterizes a diverse information offer in this particular conception of diversity. We make use of a data set of simulated article recommendations based on actual content of one of the major Dutch broadsheet newspapers and its users (N=21,973 articles, N=500 users). We find that all of the recommendation logics under study proved to lead to a rather diverse set of recommendations that are on par with human editors and that basing recommendations on user histories can substantially increase topic diversity within a recommendation set.

3bij3 – Developing a framework for researching recommender systems and their effects

“…With our suggested method, we hope to have offered an approach to scholars interested in algorithmic gatekeeping that 3BIJ3 FRAMEWORK FOR RECOMMENDER SYSTEMS 24 allows them to study its usage and effects in a more controlled environment than when studying existing black-box algorithms like Google News and similar, without compromising the realistic setting that cannot be achieved in traditional lab experiments with simulated stimuli. The same holds true for scholars interested in the interplay between algorithms and selective exposure (see Möller et al, 2018). But also framing scholars can benefit from our approach: Peperkamp and Berendt (2018) highlight that news recommender systems can influence the propagation of different frames.…”

“…We query the feeds of pre-defined sources every thirty minutes, keeping 3BIJ3 FRAMEWORK FOR RECOMMENDER SYSTEMS 12 Figure 1. Overview of the framework the selection up-to-date, and used the scrapers written for a larger project (Trilling et al, 2018) to extract the whole content of the articles, including title, teaser, text, and pictures, by parsing the HTML content that we retrieved by following the links provided by the RSS feeds. The resulting data is saved in an ElasticSearch database.…”

“…3bij3 retrieves its articles from an ElasticSearch database. We used the tool Inca (Trilling et al, 2018) to continuously populate this database with articles retrieved via a combination of RSS-feeds and web scraping, and Inca makes it easy to add any custom scraper. But also any other scraper, for any other news source, could be used, as long as the scraped data are then put into an ElasticSearch database.…”

“…Communication researchers need a better understanding of how recommender systems affect selective exposure, political attitudes, and knowledge.So far, the effect of algorithm-based selection on the diversity of news diets has mostly been discussed negatively, assuming that such systems limit the breadth of viewpoints and topics. However, recent studies challenge this conception by showing that, especially compared to other selection processes (e.g., by human editors), algorithms might not lead to more narrowed media diets after all (Möller, Trilling, Helberger, & van Es, 2018;Nguyen, Hui, Harper, Terveen, & Konstan, 2014). To provide researchers with a tool to contribute to this debate, this article sets out to develop one of the first research designs to tackle the issue of studying recommender systems in the context of news and political communication.…”

Tubes and bubbles topological confinement of YouTube recommendations

“…Contrarily, perhaps, to intuitions related to the popularization of socalled "filter bubbles", several recent studies appear to show that algorithmic suggestions do not necessarily contribute to restrict the horizon of users. Be it in terms of interaction or information consumption, users do not seem to be proposed less diversity content in regard to what would happen in the absence of recommendation [1][2][3][4][5][6] or using distinct recommendation approaches [7,8], except for what stems from explicit personalization (i.e. explicitly chosen [9], or self-selected [10], by users [11]).…”