Portuguese 
English 
Spanish 
4 min of reading 
0 comments 
The Research Analyzed How The Online Feed, Organized By Personalized Recommendation Algorithms, Reduces The Diversity Of Explored Information, Increases Confidence In Incorrect Answers, And Compromises Learning, Even When Participants Had No Prior Knowledge Of The Studied Topic.
Algorithms that organize the online feed can reduce real learning and increase confidence in mistakes, according to a study with 346 participants that analyzed how personalized recommendation systems influence the exploration of information and the understanding of new content.
Personalized Algorithms And The Direct Impact On Learning
Recommendation systems shape the online feed based on users’ past behavior, prioritizing content similar to what has already been consumed. A new study indicates that this mechanism can hinder learning by limiting the variety of information explored, even when the individual has no prior knowledge of the subject.
The researchers observed that participants who studied content selected by algorithms accessed only a fraction of the available material. Instead of exploring the entire set of information, they focused on a reduced subset, which compromised their performance in later tests.
-
Artificial intelligence is skyrocketing energy consumption, raising emissions from tech giants and pushing Google, Microsoft, and Meta closer to natural gas.
-
Mercor paid $1.5 million per day for doctors, lawyers, and former Goldman Sachs bankers to teach artificial intelligence to do their jobs, and in 17 months, it went from zero to $500 million in annual revenue while its own contractors accelerated the replacement of their own work.
-
Casio Unveils Moflin, Robotic Pet With Artificial Intelligence Designed To Provide Emotional Comfort And Simulate A Permanent Affectionate Bond
-
Google Maps Revolutionizes GPS Navigation With Gemini Artificial Intelligence, Ask Maps Feature, and Advanced 3D Visualization for Route and Travel Planning
Despite frequent mistakes, participants demonstrated a high degree of confidence in their responses. This overconfidence, associated with the restricted consumption of information in the online feed, was pointed out as one of the most concerning results of the study conducted by Giwon Bahg.
Bias Formation Even Without Prior Knowledge
Previous studies have primarily analyzed how personalized algorithms affect political or social opinions already known to users. The new research broadens this scenario by demonstrating that biases can arise immediately, even when the topic is completely unknown.
According to Bahg, algorithms can create a distorted view of reality from the first contact with the content.
Currently a postdoctoral researcher at Pennsylvania State University, he stated that personalization can induce inaccurate generalizations early in the learning process.
The research was published in the Journal of Experimental Psychology: General and involved collaboration with Brandon Turner, a psychology professor at Ohio State University.
The Effect Of The Online Feed On Perception Of Knowledge
Turner highlighted that users tend to treat limited information presented in the online feed as if it represents the complete picture of a topic.
By following only algorithmic recommendations, people believe they understand characteristics and aspects they have never explored.
This behavior leads to the loss of relevant information and amplifies the feeling of mastery over the subject.
The problem worsens when the user does not realize they are dealing with a partial selection of the available content, reinforcing erroneous conclusions.
The researchers warn that this dynamic is not restricted to specific digital contexts, but can affect different areas of everyday learning mediated by algorithms.
Practical Example Of Algorithmic Distortion
To illustrate the phenomenon, the authors presented the example of a person who decides to explore, for the first time, films produced in an unknown country. A streaming service offers initial recommendations, prioritizing a specific genre.
After choosing an action movie displayed at the top of the list, the algorithm starts recommending similar works. As a result, the user continues consuming only that genre, believing they are getting to know the cinematic landscape of the country.
This process can lead to the exclusion of acclaimed films from other genres and the formation of generalized and inaccurate ideas about the local culture. The online feed, in this case, directs the experience and limits the diversity of accessed information.
Controlled Experiment With Fictitious Learning
To test the effect of personalization, the researchers conducted an online experiment with 346 participants.
The strategy was to eliminate any prior knowledge through a fictitious learning task involving categories of imaginary aliens, similar to crystals.
Each type of alien had six distinct characteristics, varying among categories. Participants were required to learn to correctly identify these types, without knowing how many categories existed, by analyzing the available characteristics.
The experimental design allowed for the observation of how different forms of access to information influenced content exploration and participants’ ability to generalize.
When The Algorithm Guides Content Exploration
During the experiment, the characteristics of the aliens were hidden behind gray boxes. In one condition, participants needed to examine all characteristics, ensuring a complete understanding of the relationships among them.
In another condition, participants chose which resources to click on while an algorithm personalized the online feed of the study, encouraging the repetition of the same elements. Although all material remained accessible, the system reinforced previous choices.
The results showed that participants guided by the algorithm analyzed fewer characteristics and did so selectively. When tested with new examples, they made more classification errors but demonstrated higher confidence, even when wrong, evidencing a concerning pattern.
Implications For Education And Society
The conclusions raise concerns that extend beyond the experimental environment. Turner emphasized that children and young people using digital platforms to learn about the world may be affected by algorithms that prioritize engagement over informational diversity.
Repeatedly consuming similar content in the online feed does not always favor learning. According to the authors, this model can generate negative consequences for users and, on a larger scale, for society, by reinforcing limited views and excessive confidence in incomplete knowledge.
The study, titled “Algorithmic Personalization Of Information Can Cause Inaccurate Generalization And Overconfidence,” was published in September 2025 and also involved co-author Vladimir Sloutsky, a psychology professor at Ohio State University.
Seja o primeiro a reagir!