Are algorithms ruling our choices? Mechanisms, companies, and the limits of digital power

MarGib July 02, 2026

🌐 🇵🇱 Polski · 🇬🇧 EN

Artificial intelligence decides what we see on the internet, which ads reach us, and even who we vote for. But do algorithms really rule our choices — or do they merely amplify existing trends? We have analyzed the mechanisms, companies, and regulations to separate facts from myths.

Ilustracja przedstawiająca cyfrową sieć powiązań między algorytmami a ludźmi, symbolizującą wpływ sztucznej inteligencji na podejmowanie decyzji — A digital web of algorithms wraps around human silhouettes, symbolizing the growing influence of artificial intelligence on our choices and decisions

How do algorithms predict our decisions? Mechanisms that govern the digital world

Digital platforms use several key technologies to predict our behavior — both in the consumer and political spheres. However, these are not magical tools that "know" everything about us: their operation is based on specific algorithms that have their strengths and limitations.

1. Machine learning: How do they recognize our patterns?

Machine learning (ML) models analyze massive datasets — browsing history, purchases, social interactions — to identify patterns. For example, the YouTube recommendation system (based on deep neural networks) processes over 80 miliardów zdarzeń dziennie to suggest the next videos. It is no coincidence that these match our tastes: the algorithm learns based on our previous choices and the behavior of similar users.

In marketing and politics, this technique is used for message personalization. Tools like Cambridge Analytica (known for the controversy surrounding the 2016 US election) used psychometric models, such as OCEAN classification, to tailor advertising content to the individual personality traits of users.

2. Reinforcement learning: How platforms adapt to us in real-time

Reinforcement learning (RL) algorithms are used primarily by e-commerce platforms and social media. They operate on the principle of continuous testing and optimization: if we see an ad and click on it, the system considers this a "positive reinforcement" and will display it to us more often. This is how Amazon Personalize works, adjusting offers based on our current actions.

In a political context, RL can be used for dynamic message shaping depending on user reactions. For example, political campaigns can modify advertising content in real-time based on sentiment analysis on social media — for instance, using tools like Brandwatch, which monitor discussions on Twitter.

3. NLP: How algorithms read our thoughts (or at least try to)

Natural Language Processing (NLP) allows for the analysis of massive online discussions to draw conclusions about social sentiment. Platforms like Twitter or Facebook use NLP to identify key topics and the emotions associated with them — for example, before an election. Such tools help political parties tailor their message, but they can also reinforce polarization by promoting content that triggers strong emotional reactions.

Who rules the algorithms? Companies shaping digital reality

The influence of algorithms on our decisions would not be possible without the specific entities that develop and implement them. Some are global corporations, while others are specialized firms operating in the shadows.

1. Tech giants: Meta, Google, Amazon

The leading players are primarily:

Google (YouTube, Google Ads) — the YouTube recommendation system was described in a 2018 study as one of the most influential on user decisions, even shaping their political views (ACM, 2018).
Amazon — uses machine learning to personalize shopping, which influences not only consumption but also how consumers perceive their needs.

2. Political microtargeting specialists

Beyond corporations, there are firms that specialize in influencing political decisions through advanced analytical tools:

i360 (USA) — provides microtargeting tools for political parties, including in Donald Trump's 2020 campaign.
Predictive Solutions (Poland) — uses predictive models in political marketing, helping parties reach specific groups of voters.
Defunct firms, such as Cambridge Analytica, which disappeared from the market after the scandal, but whose technologies are still being developed by other entities.

Does the law keep up with the digital revolution? Regulations that should protect us from algorithms

Although algorithms have a huge impact on our lives, legal systems are still trying to get a handle on them. Some regulations work, but many loopholes remain unfilled.

1. GDPR and data protection: The gold standard with loopholes

GDPR introduced the obligation of transparency in personal data processing and gave users the right to object to automated decision-making (e.g., personalized ads). In practice, however:

Companies often interpret GDPR to their advantage, e.g., through complex privacy policies.
Fines are rare and disproportionate to the profits corporations make from using data. An example is the 20 million euro fine imposed on Amazon by CNIL (France) in 2021 for non-compliance with GDPR in advertising.

2. National laws: Poland between protection and helplessness

In Poland, the President of the Personal Data Protection Office (PUODO) conducts audits of companies violating regulations. In 2022, a fine of 1.2 million PLN was imposed on Allegro for improper processing of customer data. However:

The ban on conducting election campaigns using personal data without the voter's consent (Art. 162 of the Election Act) is difficult to enforce against foreign platforms like Facebook.
There is a lack of specific regulations regarding artificial intelligence in a political context. The AI Act (European Union) is intended to introduce restrictions, but it has not yet been fully implemented.

3. Why are regulations insufficient?

Current regulations focus mainly on data protection, not on the influence of algorithms on democratic processes. In the US, there is no federal law limiting political microtargeting — only the state of California has introduced the California Consumer Privacy Act (CCPA), which has limited application.

Do algorithms decide elections? Facts and documented cases

One thing is certain: algorithms influence what information reaches us and how we perceive it. But have they decided the outcome of any election on their own? It is difficult to give a definitive answer to this question.

1. US Presidential Election (2016) and Brexit Referendum (2016)

In both cases, companies like Facebook and Cambridge Analytica were accused of using microtargeting to influence voters. Reports indicate that:

Algorithms targeted users with specific political preferences, reinforcing existing social divisions (The Guardian, 2018).
In the case of Brexit, tools like AggregateIQ used psychographic data to create personalized political ads (The Observer, 2017).

But: There is no evidence that algorithms decided the election results on their own. Their influence mainly consisted of reinforcing existing polarization and shaping the public agenda.

2. Elections in Brazil (2018) and India (2019)

In Brazil, algorithms on WhatsApp and Facebook contributed to the spread of false information that influenced voter sentiment (BBC, 2018). In India, algorithms amplified polarizing content, which may have influenced the voting outcome (Reuters, 2019).

3. Elections in Poland (2023)

In Poland, political parties, including Law and Justice (PiS), used advanced ad targeting on Facebook. According to OKO.press reports, these tools allowed for very precise reaching of specific voter groups, e.g., through ads aimed at young men from specific regions (OKO.press, 2023).

Conclusions: Algorithms do not "lose" elections on their own, but they can change their course, especially among narrow, key voter groups. A study published in Science Advances in 2020 found that microtargeting on Facebook could have increased voter turnout by 3–5% among specific groups (Science Advances, 2020).

Are algorithms the new power? Experts debate the definition

The question of whether algorithms constitute a "new power" is the subject of lively debate among scientists, ethicists, and politicians. Some see them as a threat to democracy, others as merely a tool whose influence can be limited.

1. Arguments that algorithms are a new form of power

Proponents of this thesis point to several key points:

Discretionary power: Algorithms decide what we see on the internet, shaping our reality. For example, the YouTube recommendation system steers users toward increasingly extreme content, which can lead to radicalization (The New York Times, 2019).

2. Arguments against the "new power" thesis

Skeptics point to the limitations of algorithms and the possibilities of countering their influence:

Lack of sovereignty: Algorithms operate within existing legal frameworks, although these are often insufficient. In 2023, the Court of Justice of the EU ruled that Facebook's algorithms must comply with European law (Curia, 2023).
Technical limitations: Algorithms are not perfect. For example, a 2021 MIT study found that facial recognition algorithms have significantly lower accuracy for people with darker skin (MIT News, 2021).
Possibility of counteraction: Users can change privacy settings or use alternative platforms. For example, the #deletefacebook movement in 2018 led to a partial drop in users (TechCrunch, 2018).

3. Expert perspectives

In summary, experts are divided:

Evgeny Morozov (media theorist): "Algorithms are a new form of power, but they are neither infallible nor uncontrolled."

Yuval Noah Harari (historian): "Tech companies are becoming new empires, but their power is not yet fully established."

What are the limits of algorithms? Why are they not infallible?

Although algorithms can appear "all-knowing," their operation has serious limitations. Knowledge about them is key to not overestimating their influence.

1. The problem of bias in data

Algorithms often repeat human biases present in training data. For example, the COMPAS system (used in the US to assess recidivism risk) was accused of racial discrimination because it repeated judges' biases against people with darker skin (ProPublica, 2016).

2. Low-quality data leads to erroneous predictions

Errors in input data are common and lead to incorrect results. A 2022 Stanford University study found that predictive models in medicine often fail due to low-quality data (Stanford Medicine, 2022).

3. The volatility of human behavior

People make decisions under the influence of emotions, which are difficult to predict. A study published in the Harvard Business Review in 2021 showed that algorithms have 30% lower accuracy in predicting purchasing decisions compared to logical decisions (HBR, 2021).

What does science say? Studies that confirm and debunk the influence of algorithms

To separate facts from myths, it is worth looking at the results of independent scientific studies.

1. Confirming the influence of algorithms on social polarization

Studies conducted by the Oxford Internet Institute in 2020 showed that social media algorithms reinforce political polarization, but do not decide election outcomes (Oxford Internet Institute, 2020).

2. YouTube algorithms accelerate radicalization

A study conducted by MIT and CNRS in 2022 proved that the YouTube recommendation system accelerates the radicalization of users by recommending increasingly extreme content (MIT News, 2022).

3. Political microtargeting is effective, but its influence is often overstated

A study published in the Journal of Economic Perspectives in 2021 found that political microtargeting is effective, but its influence on election results is often overstated by the media (AEWeb, 2021).

4. Most algorithms have systemic biases

The AI Now Institute report from 2023 confirmed that most predictive algorithms have systemic biases that affect their accuracy (AI Now Institute, 2023).

Summary: What do we really know about algorithms and our choices?

Algorithms have a huge impact on what information reaches us, what decisions we make, and how we perceive the world. However, they are not omnipotent. Their operation is based on specific mechanisms that have both strengths and serious limitations.

Key takeaways:

Algorithms reinforce existing trends (e.g., social polarization), but rarely decide election outcomes.
Their influence is greatest among narrow, key voter groups, such as young people or residents of specific regions.
Current regulations are insufficient to effectively limit their influence on democratic processes.
Technical limitations of algorithms (e.g., data bias, low data quality) mean their accuracy is much lower than commonly believed.

What can we do? How to limit the influence of algorithms on our lives?

Although we cannot completely escape the influence of algorithms, we can take several steps to minimize their impact on our decisions:

Change privacy settings on social media and browsers — limit data collection.
Use alternative platforms — e.g., less commercialized search engines or social media.
Be a conscious user — ask yourself why a piece of content was displayed to you and whether it is objective.
Support regulations limiting the influence of algorithms — e.g., by participating in public consultations or voting for politicians who take this problem seriously.

Algorithms are not all-powerful, but their influence on our lives will only increase. Knowing how they work is the key to making more informed decisions — both in the political and consumer spheres.