Many online platforms – e.g., video portals, e-commerce, social networks – use recommendation algorithms to recommend products – e.g., videos, articles for sale, or news – to their users as accurately as possible. Such recommendations are based on various principles. For example, what can be recommended is what many other users with similar interests also like or what is similar to the previous products of the respective user. However, products for which a provider pays particularly high advertising prices can also be recommended.
The simulator „Plattformdynamiken“ simulates the recommendation system of a video portal. By changing various parameters, the principles implemented in the simulator can be explored. The perspective of the users with their respective content interests is included as well as the perspective of the platform operators, who pursue financial interests, among other things.
Recommendation algorithms are usually black boxes, which means that their exact function is not disclosed. For example, it remains intransparent to what extent the users‘ behavior influences the recommendations or on which assumptions the recommendation algorithms are based – e.g., those that can also lead to the reinforcement of biases. Even an analysis of user behavior or user data cannot (fully) clarify these questions.
Using the simulator allows the discovery of features that give rise to discussion and comparison with existing platforms. However, it also contains some ambiguities and aspects that are difficult to understand and illustrate the challenges in comprehending recommendation algorithms. The simulator is particularly suitable for media education at school as well as for general media education. The „Plattformdynamiken“ application was created as part of the DataSkop project.
This project was bootstrapped with Create React App.
MIT