The invisible architect: algorithms in social media

In the contemporary digital age, social media platforms have become central to how individuals communicate, consume information and form opinions. At the core of these platforms lie algorithms, invisible yet powerful systems that organise, filter and prioritise content for billions of users worldwide.

In the context of social media, algorithms decide which posts, videos and trends users see by analysing data such as likes, shares, watch time, search history and interaction patterns. These systems use machine learning models to predict what content will keep users engaged for longer, becoming more accurate as more data is collected. However, they often prioritise engagement and relevance over accuracy, social value or ethical concerns.

The personalisation created by algorithms has transformed how individuals experience reality online. Users are increasingly exposed to content that aligns with their existing beliefs and match their interests. This can contribute to phenomenon referred as "filter bubbles" or "echo chambers" although these are not produced by algorithms alone. Key findings indicate that algorithmic feeds increase user engagement roughly by 15-30% but reduce content diversity and elevate polarising material.

Importantly, these algorithms do not operate neutrally; they function selectively, reflecting the strategic interests and policy priorities of the actors or states that influence them. This becomes evident in practices such as the selective banning of certain platforms in specific countries while allowing them to operate freely in others.

Similarly, the politicisation of social media spaces varies by region, where narratives are amplified or suppressed depending on strategic alignments. For instance, X activated a crisis misinformation policy for the 2020 US election and the Russia-Ukraine war, but did not actuate the same level of moderation protocols for the India-Pakistan border escalation in 2025 or the Myanmar coup in 2021, despite comparable levels of misinformation. This suggests that uneven enforcement was not produced by algorithmic amplification alone, but from their interaction with selective policy interventions.

Globally, countries have adopted different strategies to address this regulation concern and a key trend across all models is the evolving role of regulators. Regulatory systems are evolving as existing bodies take on wider roles and new institutions are created to specifically to deal with algorithmic influence. This reflects a broader move towards a state-based governance system.

The US largely relies on a market-driven model. Largely shielded by ''Section 230 of the Communications Decency Act'', digital platforms are generally protected from liability for user-generated content. However, concerns over privacy, misinformation, alleged political bias and harmful platform design have increased pressure for reforms, leading to a reconsideration, with proposed bills like the "Filter Bubble Transparency Act" seeking to mandate algorithmic choice.

Conversely, the EU has taken a more interventionist stance. It has positioned itself as the global standard bearer for digital rights. The General Data Protection Regulation (GDPR) limits how user data can be harvested. More recently, the Digital Services Act (DSA) and Digital Markets Act (DMA) directly target algorithmic transparency. Under the DSA, very large online platforms must assess the risks of their systems and offer users the option of a feed that is not based on profiling (a non-algorithmic feed).

In China, the state exercises direct control over algorithmic systems, requiring platforms to align with national priorities and social norms.

However, the creation and control of these powerful algorithms have become another central battleground between the large global tech companies that build them and the state governance systems that seek to manage them. The data highlights that the Meta Platforms annual revenue for 2024 was $164.501B, a 21.94% increase from 2023 that must be protected as a trade secret. Any mandated change that could reduce engagement by even 1-2% threatens billions in revenue. In this regard, the economic asymmetry of Pakistan is overwhelming as Pakistan's entire federal budget (2023-2024) was $51 billion smaller than what Meta Platforms earns in a year. This illustrates the disparity in negotiating over algorithmic governance as states seek sovereignty over information spaces, while companies protect proprietary algorithms as commercial assets.

Pakistan is one of the fastest-growing digital market (the 10th largest in the world) with about 116 million internet users, and 71 million active social media networks (SMNs) users, yet its engagement with algorithmic governance remains limited. SMNs in Pakistan are primarily controlled and regulated by Pakistan Telecommunication Authority (PTA), which operates under the section 37 of Prevention of Electronic Crimes Act (PECA) 2016. As of March 2026, the government has formally activated a new regulatory body, the Social Media Protection and Regulatory Authority (SMPRA) to handle fast-track content regulation and complaints about misinformation, besides ensuring that platforms comply with registration, data localisation and content removal within 24–48 hours.

Therefore, a key challenge in Pakistan is the limited regulatory capacity. Institutions such as PTA and SMPRA are limited in their ability to influence platform behaviour, as they cannot control algorithms directly or enforce compliance beyond extreme measures such as blocking excess and content removal.

Another issue is the data sovereignty and control as majority of Pakistan's digital public sphere operates on foreign-owned platforms such as Facebook, Instagram and WhatsApp, all controlled by Meta Platforms. Thus, decisions about content ranking, visibility and moderation are made outside Pakistan's legal and cultural framework.

Public awareness of algorithms in Pakistan is also very low. Many users perceive their social media feeds as neutral reflections of reality, rather than curated outputs shaped by engagement-driven algorithms. This limited awareness is compounded by the underdevelopment of academic research, institutional capacity and policy expertise on algorithmic governance.

Pakistan faces a policy dilemma between ensuring national security and protecting freedom of expression. Balancing these competing priorities remains one of the biggest challenges in governing algorithms effectively. Therefore, a fundamental shift is required to understand how Pakistan conceptualises and engages with global tech firms. These platforms must be treated not merely as commercial entities but as transnational power centres. Thus, policy engagement with them should move beyond a regulator to company approach and adopt a state-to-state or state-to-global actor framework, recognising their geopolitical significance.

In nutshell, given the growing challenges of algorithmic governance and its implications for national security, the core lesson for Pakistan lies in shifting its regulatory focus from reactive content control to proactive governance. The path forward requires a whole of government approach, strengthening institutional capacity to engage with platforms on issues of transparency on algorithmic impact, and improving public awareness of how content is curated and amplified.