Algorithmic personalization study. Who and how understands digital media

illustration: DALL-E

In the age of digital dominance, where every click and swipe leaves a trace, the topic of content personalization takes on a new dimension. Vaclav Moravec from Charles University in Prague and a team of researchers from the Czech Republic, Croatia, Poland, Slovakia, and Ukraine examined how different social groups understand the phenomenon of online content personalization. Based on data collected from 1,213 Czech citizens, the authors show that awareness of how personalization algorithms work in media is highly socially differentiated. The results of the study were published in the article “Algorithmic personalization: a study of knowledge gaps and digital media literacy” in Humanities and Social Sciences Communications, part of the Nature Portfolio.

Content personalization: opportunity or threat?

The main goal of the study was to analyze citizens’ knowledge of how online services tailor content to users. Personalization relies on data - from search history to location and shopping preferences - and serves to increase message relevance. But while the benefits are clear (faster access to relevant information, tailored ads), there is growing concern about the loss of privacy and informational manipulation.


The authors of the “Algorithmic personalization” report emphasize that understanding personalization mechanisms is key not only to protecting personal data but also to the ability to critically assess content. Informed individuals can influence algorithms, limit manipulation, and better protect their privacy.

Who knows more? Social differences in knowledge about algorithms

Researchers used a three-step analysis system based on an informational model, fuzzy logic method, and social classification. They asked respondents about:

• awareness that online content is personalized,
• knowledge of technical ways content is customized,
• feeling of control over what they see online.

Based on these three areas, the researchers developed knowledge metrics that include both objective understanding of how algorithms work and users’ subjective sense of influence over their digital environment. The study`s authors used data from 1,213 surveys, analyzing them with demographic variables such as age, gender, and education level. The results made it possible to define knowledge levels about personalization for different social groups with numerical values between 0 and 1 - the higher the value, the greater the user’s awareness.

Social group	Knowledge level about personalization
Men aged 35-44, higher education	0.812 (above average)
Women aged 15-24, higher education	0.821-0.812
People with vocational education	0.661-0.672 (low scores)

source: “Algorithmic personalization: a study of knowledge gaps and digital media literacy”

These differences show that younger users with higher education better understand how content-personalizing algorithms work. Meanwhile, those with lower education levels more often report a lack of control over what they see online. This raises questions about informational equality and the need for education.

Real-life example. Netflix and the illusion of choice

Let’s imagine two Netflix users - a thirty-year-old woman with higher education and a fifty-year-old manual worker. Although both open the same platform, the algorithm shows them completely different recommendations. She sees political documentaries and indie films, while he gets comedies and entertainment shows. Both believe they have a choice, but in reality, they watch what the algorithm deems suitable. If they are unaware of this mechanism, they may wrongly assume it’s the result of their “free will.”

This is exactly the issue raised in the report by Moravec and his co-authors - the invisible hand of the algorithm guides the user, who often doesn’t realize it.

Education as the answer

The research team proposes concrete solutions. Their analytical-information system allows for nuanced evaluation of citizens’ knowledge, accounting for demographic differences. Notably:

• the system works regardless of the number of criteria or questions,
• it can be adapted to other regions or countries,
• it enables the implementation of educational programs tailored to specific social groups.

Thanks to this tool, it is possible, for example, to identify groups most vulnerable to manipulation and include them in specialized media education programs. As the “Algorithmic personalization” report shows, such actions are necessary - especially in the face of the risk that personalization deepens social divides, traps users in information bubbles, and facilitates the spread of disinformation.

Among the researchers’ plans is the development of software that will allow practical application of the methodology in different contexts. The next step is to analyze factors influencing the acceptance of personalized content and to study the relationships between perceived disinformation risks and trust in media.

The study by Moravec and his co-authors is not only a scientific analysis but also a call to action for policymakers - from NGOs to state institutions - not to ignore differences in digital competencies. The more people understand how algorithms work, the lower the risk that society will become a passive recipient of manufactured content.

* * *
More about the study: Moravec V., Hynek N., Skare M., Gavurova B., Polishchuk V. (2025). Algorithmic personalization: a study of knowledge gaps and digital media literacy. Humanities and Social Sciences Communications. https://doi.org/10.1057/s41599-025-04593-6