DREAM / The Prime Minister resigns and wakes up under the Parisian bridge!

 Kliko këtu për të tjera
Voxpress:

The urgency of reforming algorithms in the AI ​​era

2025-07-08 12:50:57Pikëpamje SHKRUAR NGA EMIRJON SENJA
The urgency of reforming algorithms in the AI ​​era
Emirjon Senja

The enthusiasm for using algorithms to predict citizen behavior in order to prevent deviant events, as well as in the process of distributing content on social networks, has now ended.

Time has shown that in many cases, when used for these purposes, algorithms failed, showing pronounced elements of social profiling and even racism, while as for social networks, it has already been proven that they favor partisanship instead of a healthy clash of ideas, polarizing societies through the reinforcement of prejudices.

Algorithms and AI as drivers of polarization in public discourse

When consuming content on social networks, we are not the ones who select what we will see on our phone or computer screen, but rather the algorithms that decide what will be shown to us depending on our online behavior as well as the way the algorithm of the social network we are on is programmed, playing an important role in shaping the way users perceive reality.

Programmed to increase user interaction and keep them on the platform as long as possible, an approach through which they also determine the flow of information on social networks, algorithms have created a fragmented and polarized ecosystem of online discourse.

Social networks like META; X; Youtube, TikTok have programmed their algorithms to personalize the content they display to users, which is based on their previous online behavior, as well as their circle of followers or followers on the platform, with the ultimate goal of keeping their audience there longer and increasing engagement.

Various studies have confirmed that this model feeds a cycle where users are continuously exposed to content that reinforces their beliefs, due to interaction with the same content, limiting them to other perspectives.

As a result, what experts in the field call "echo chambers" and "information bubbles" are formed, where the user is "bombarded" with the same information, a phenomenon that encourages the radicalization of discourse and deepening divisions of attitudes in society.

In a global context with social and political tensions, as well as armed conflicts in different regions, the way social media algorithms work, instead of helping to provide alternative thinking, is further reinforcing users' initial ideas, pushing their positions even further towards extremes.

Consequently, today, social networks are not uniting, but rather dividing societies into islands of belief, where everyone there finds themselves deprived of other ideas.

Koo, GH, Johnson, TJ, Santillana, M., & Jeon, G[1]. study on the toxicity of online political participation highlights how the interaction of algorithms with users' emotional and political behaviors reinforces opposing positions. In this context, the Artificial Intelligence platform used to personalize content and promote it has shown itself to be not neutral.

Instead, AI automatically learns and then promotes behavioral patterns considered successful in terms of increasing interaction on social networks.

Social networks flooded with AI content

The rise of Artificial Intelligence platforms for video production; the reduction in subscription costs due to strong competition in this sector, as well as machine learning with the Albanian language, are confronting social media users with the great risk of perceiving a false reality on social media.

In recent months, Albanian social media users have been increasingly interacting with AI-generated video content. They are without any relevant annotations from publishers, which, due to their provocative nature, are being offered as Vox Pop interviews, as natural responses from citizens, and are generating a large number of interactions.

A combination of the power of algorithms to effectively target users in the service of increasing interactions, as well as content that is easily confused with reality produced by Artificial Intelligence, has become a major risk for social media users.

According to data published by Smart Insights[2] in February 2025, a video generates 5 times more interactions compared to other content published on social networks due to its dynamic nature.

Another study[3] shows that users spend an average of 100 minutes a day watching videos online, which confirms the dominance of the video format over others.

Data shows that on platforms like Facebook and Youtube, the average age of users in Albania increases year by year, while the situation becomes even more problematic given the low level of media literacy in a society like ours.

The impact of AI on content personalization and the promotion of online radicalization

Despite having the same goal, namely increasing engagement, social networks apply different targeting practices, which essentially produce personalized information on each user's "wall".

Personalizing information is now standard practice on social media platforms, creating a fragmented information reality for users due to the lack of alternative ideas.

A study by the European Parliamentary Research Office in 202[4] 4 highlights that algorithmic personalization directly affects the way individuals consume political information, significantly increasing the risk of their being misinformed and the spread of conspiracy theories.

The high data processing capabilities of Artificial Intelligence further fragment information for users. This is also noticeable in the way META and TikTok algorithms, for example, distribute information across platforms.

META's algorithms distribute information based on a user's connections and contacts on Facebook or Instagram, believing that people interact more with content shared by those they know or trust, while TikTok uses a completely different practice.

TikTok is based not on social connections, but on the audience's interest in the content's topic. Thus, TikTok's algorithms can display content even from profiles considered "foreign", but that match the topic that the user has shown interest in while browsing the Internet.

Both practices lead to informational isolation of the user, although in the case of TikTok, due to the very way its algorithms operate, the risk is greater.

It's time to reform algorithms

The way in which much of the content is being produced with Artificial Intelligence and how online content is being distributed through increasingly powerful algorithms makes it necessary to review the model being used by networking platforms to target audiences.

Various researchers in the field suggest a fundamental reform of the approach first and then of the way algorithms behave during the process of distributing content on social networks.

One of the most talked-about approaches is that of two Oxford University professors Sandra Wachter and Dr Brent Mittelstadt[5], who suggest developing “fair distribution” algorithms that provide ideological balance in content, while also exposing users to opposing viewpoints. They call this “human rights-compliant algorithmic justice.”

It is proposed that algorithmic systems that serve to distribute content should not only consider interaction as highly efficient, but during this process, take into account freedom of expression, equality in access to information, and above all: the right to speech for all.

They suggest a certification of AI algorithms and platforms by independent institutions before their introduction into use. A similar suggestion was made by Yuval Noah Harari in 2023[6].

It is also suggested to implement transparency mechanisms on how algorithms work.

The role of Artificial Intelligence in this new context can be transformed from an amplifier of polarization in society, to a promoter of dialogue and the clash of different opinions on a platform, which now behaves “honestly” with users. This can happen if the algorithms themselves, but also the AI, are trained on the principles of diversity of opinions and social stability.

For example, instead of promoting extreme positions, which lead to more interactions, AI could identify and promote content that fosters social cohesion and constructive dialogue by highlighting marginalized voices in public discourse. The narrative would be completely different.

Another suggested model is that of “diversity of exposure”[7], which is based on the principle of intentionally showing users content with perspectives different from their own.

Other researchers[8] go even further by suggesting a model of algorithmic transparency, regulated through independent institutions based on the argument that platforms cannot self-regulate and that public oversight is needed to ensure that distribution mechanisms do not promote extremism.

Their stances come in a problematic context for how algorithms powered by the vast processing capacities of AI work today, yet they conflict with a general approach in the information technology sector, which is more based on, or prefers, self-regulation.

Essentially, the need to rethink how algorithms guide the distribution of information on social media platforms driven by Artificial Intelligence is no longer seen as just an ethical or technical issue, but a social necessity in defense of democracy and social cohesion.

However, the necessity to review algorithms expressed by experts in the field conflicts with the business model of social media platforms, which itself has produced the logic of how algorithms operate, whose effectiveness is measured by the high number of interactions on the platform.

Experience to date has shown that larger interactions result in polarizing content.

This is the moment where either societies will change algorithms, or the latter will completely take control of public discourse, ultimately changing societies.


[1] https://www.tandfonline.com/doi/abs/10.1080/19331681.2025.2504518

[2] https://www.smartinsights.com/social-media-marketing/social-media-strategy/new-global-social-media-research/?utm_source

[3] https://www.charleagency.com/articles/video-marketing-statistics/?utm_

[4] https://www.europarl.europa.eu/RegData/etudes/STUD/2022/729541/EPRS_STU(2022)729541_EN.pdf

[5] https://committees.parliament.uk/writtenevidence/42151/pdf/?utm_source

[6] https://www.economist.com/by-invitation/2023/04/28/yuval-noah-harari-argues-that-ai-has-hacked-the-operating-system-of-human-civilisation

[7] https://academic.oup.com/pnasnexus/article/3/12/pgae518/7904735

[8] https://jolt.law.harvard.edu/assets/articlePDFs/v372/3-The-Meta-Oversight-Board-and-the-Empty-Promise-of-Legitimacy.pdf?


Në përputhje me standarded JTI

Video

SHIKO KANALIN NE YOUTUBE
×