Algorithmic recommendation systems. What does the German public think about the use and design of algorithmic recommendation systems?

Algorithmic recommendation systems are regularly used by a majority of the population. The recommendations given are usually based on large amounts of data collected about users. The evaluation of the data takes place both on a supervised basis and as part of a self-learning process. Research on the so-called automation bias assumes that people tend to follow recommendations made by algorithms. Even if, for example, they merely prepare people’s consumption decisions, they come quite close to being an automated decision-making system in this respect. However, it is unclear how the German population think about such systems: What are the opinions on the consequences of algorithmic recommendations? And based on which data are respondents more likely to opt for the best possible outcome? Our data from the Opinion Monitor Artificial Intelligence (Meinungsmonitor KI [MeMo:KI]) show, that in many application areas (e.g., on music platforms or in media libraries), algorithmic recommendation systems are perceived as useful. However, a closer look paradoxically reveals that many respondents expect only limited time savings, orientation or the best possible result from the use of such systems. Furthermore, 67 percent of respondents consider algorithmic recommendation systems to be not at all or only slightly trustworthy. Unsurprisingly, the respondents are very critical of the use of personal data by such systems, especially when it comes to information about personal contacts or consumer behavior.

Recommended citation: Kieslich, K., Došenović, P., & Marcinkowski, F. (2021). Algorithmic recommendation systems. What does the German public think about the use and design of algorithmic recommendation systems?. Factsheet No. 4 of the Meinungsmonitor Künstliche Intelligenz [Opinion Monitor Artificial Intelligence].
Download Paper