More knowledge, but still no clear standpoint – Germany’s evolving relationship with algorithms and AI

Algorithms and artificial intelligence (AI) are increasingly shaping our everyday lives. A new survey shows what the German public knows about these technologies, and what it thinks of them. A comparison of the recent results with those of 2018 and 2020 shows that while the awareness and acceptance of automated decision-making has grown in recent years, large segments of the population find it difficult to appreciate the implications of its application.

Daycare slots are allocated with the help of algorithms, people are identified in video surveillance systems in public spaces, and natural disasters are predicted in early warning systems: algorithms and artificial intelligence have come to stay. But what do people in Germany actually know about these technologies, and how do they perceive the impact of them on their daily lives and society?

Having explored these questions in representative surveys conducted in 2018 and 2020 by the Institut für Demoskopie Allensbach (IfD) on behalf of the Bertelsmann Stiftung, our recently published follow-up survey focuses on updating the trends identified. Our goal was to determine whether respondents’ self-assessed knowledge of algorithms had changed and how their attitudes about the impact of the technologies had evolved over the past four years.  The survey also explored for the first time the extent to which the term “artificial intelligence” elicited different results than the term “algorithm”.

The German public reports having more knowledge of how algorithms work and what their implications are

Over the past four years, awareness of the term “algorithm” has increased from 72% to 81%. Respondents also report that their knowledge of algorithms has increased. For example, 60% now say they have a rough or even fairly sound understanding of algorithms.

Uneven distribution of knowledge about the terms “algorithm” and “artificial intelligence”

While 81% of respondents said that they had heard the term “algorithm” before, the figure for “artificial intelligence” was slightly higher at 88%. A closer look at these figures reveals a much greater familiarity with the term “artificial intelligence” than with the term “algorithm” among older cohorts in particular. While 69% of the over-60-year-old respondents said they were familiar with the term “algorithm”, 82% in the same age group reported being familiar with the term “artificial intelligence”.

An analysis of responses sorted by educational attainment levels shows even larger gaps in this regard. Knowledge of the two terms increases with the level of formal education achieved. With regards to the term “algorithm”, we see a significantly wider gap in the level of knowledge between lower and higher educational than we do for the term “artificial intelligence.” Whereas 75% of respondents with a lower-level secondary education attainment level said they were familiar with the term “artificial intelligence,” only 54% said the same about the term “algorithm.”

At the same time, the “education gap” has widened since 2018. The share of respondents in Germany with a relatively high level of educational attainment (Abitur/university studies) who know little or nothing about the term “algorithm” has nearly halved within four years (from 31% to 17%). Among those with a low level of educational attainment (lower-level secondary school), there was a decrease of only seven percentage points to 68%.

Growing acceptance of (partially) automated decisions

Drawing on specific examples of use cases, the survey sought to determine whether people are aware of the areas in which algorithms and AI are applied. Since 2018, knowledge of algorithms and AI in the use cases queried in both years has increased by an average of more than ten percentage points. The best-known use cases were personalized advertising technology (69%), facial recognition technology in public spaces (67%), and spell/grammar-check technology in word processing programs (66%).

The acceptance of algorithms and artificial intelligence is closely linked to how familiar respondents are with the specific context in which the technologies are applied. Automated decisions are generally more acceptable the more familiar respondents are with the area in which the technology is being applied.

In addition, the findings show that people generally hold fewer reservations about (partially) automated decision-making in areas that have a comparatively low social impact. We nonetheless observe a surprisingly high acceptance rate for (partially) automated decision-making in specific applications with a strong potential for social harm. This applies, for example, to credit scoring procedures, the preselection of candidates in hiring processes and, in a particularly striking finding, the use of facial recognition technology in public spaces, the acceptance of which increased from 36% in 2018 to 51% in 2022.

Little awareness of how algorithms and artificial intelligence impact one’s own life

Remarkably, the German population still finds it difficult to recognize the relevance of algorithms in their own lives and in everyday life more broadly. In this regard, little has changed since 2018. The share of respondents who believe algorithms have a strong or very strong impact on their lives in general and everyday life more broadly has risen only slightly from 27% to 29%. In the case of AI, only 14% currently believe this to be the case.

Another relatively consistent finding is the widespread ambivalence regarding the impact of these digital technologies on individual and societal life, and whether their effects are more negative or positive. More than 40% of those polled still do not have a clear opinion on whether the growing use of algorithms and artificial intelligence will create more benefits or more risks for society as a whole. Overall, people in Germany have yet to develop a clear standpoint regarding algorithms and artificial intelligence.

Broad-based skills-building efforts, nuanced reporting, and more common good-oriented use of algorithms are needed

The survey results clearly show that much still needs to be done to ensure an informed public and to narrow the knowledge gap with regard to algorithms and artificial intelligence.

  1. Promote literacy regarding algorithms and AI among all groups in society

While increasingly fewer people report a lack of knowledge regarding algorithms, how they function and where they are used, this is not the case equally across all demographics. Respondents with a higher level of educational attainment in particular know more about algorithms and artificial intelligence. However, since algorithms and artificial intelligence influence every individual’s life and their opportunities for participating in society, we must strengthen literacy regarding these technologies, regardless of a person’s age, income or educational attainment level.

  1. Facilitate differentiated media reports aimed at revealing the often unknown applications and “hidden” effects of algorithms and AI

This is important because most people in Germany still find it extremely difficult to assess the impact these technologies are having on their lives. In addition, large parts of the population remain unaware of many areas in which algorithms and artificial intelligence are applied. In order to foster an informed public able to assess the opportunities and risks involved with these technologies, we need a more diverse media discourse that explores not only the economic, but also the social and societal effects of automated decision-making.

  1. Use algorithms and AI more often to target social issues and the challenges facing society

This is important if we are to effectively address the German public’s considerable skepticism regarding whether the use of these technologies will deliver more benefits or problems for people as individuals and society as a whole. We therefore need more examples of how algorithms and artificial intelligence can be used – particularly in everyday life – to promote the common good.

The raw data, indicators processed for statistical analysis, and the log file with the outputs of our calculations are available at: 

This text is licensed under a Creative Commons Attribution 4.0 International License