DataSkop: simulating the dynamics of recommender systems
Several thousands of supporters donated their YouTube profile data on our platform DataSkop. We have developed a simulator that is designed to demonstrate basic principles of recommendation algorithms such as YouTube's.
Recommendation systems (aka recommender systems) are deployed to predict users' interests from a pool of content and to recommend content according to these predictions. YouTube and Netflix provide typical examples for such a recommendation service.
DataSkop is a collaborative project involving AlgorithmWatch, scientists from University Viadrina Frankfurt, Paderborn University, the University of Potsdam, and mediale pfade, an association for media education. It was developed to help users having a better understanding of applications that are based on data. This includes YouTube’s collecting of their data and processing it for recommendations.
The recommender simulator “Plattformdynamiken”
Media education is an integral component of the DataSkop project. We must find suitable measures to explain the functioning and the importance of algorithmic decision-making systems, not only to children and adolescents (for whom digital applications are an indispensable part of their life) but also to adults. To this end, we developed a simulator that is designed to playfully demonstrate basic principles and dynamics of recommendation algorithms.
Check out our interactive application: https://dataskop.net/recommender-sim/?en
YouTube’s choice: background information on the data donation experiment
In a detailed report, Peter Kahlert from the European New School for Digital Studies at University Viadrina Frankfurt (Oder)–one of DataSksop project partners–describes how the donated data contributes to research. The analysis of data donations helps understanding the premises of platform dynamics as well as examining the use of platforms and the according user experience critically. DataSkop is also exploring the method of data donation itself. It is a pilot project to identify its potential and its operational limits which is important in order to compare data donation to similar research methods.
Another important aspect of DataSkop is that it promotes transparency towards third parties: civil society organizations, scientists, or the media. We need this transparency to ensure fair and inclusively designed access to the influential and ever-changing platforms. All users must be empowered to interact with platforms autonomously while being aware of their interactions’ consequences.
Read the article here: https://medium.com/sts-ens/youtube-recommends-148b5608635a