Artificial Intelligence promises unprecedented access to the world’s knowledge, yet delivers a curated illusion. As algorithms prioritise engagement over understanding, what appears open is in fact tightly controlled. Soumi Banerjee explores how algorithmic mediation deepens inequalities, shaping not just what we see — but how, and whether, we understand it
Artificial intelligence (AI) has made global movements, testimonies, and critiques seem just a swipe away. The digital realm, powered by machine learning and algorithmic recommendation systems, offers an abundance of visual, textual, and auditory information. With a few swipes or keystrokes, the unbounded world lies open before us. Yet this 'openness' conceals a fundamental paradox: the distinction between availability and accessibility.
What is technically available is not always epistemically accessible. What appears global is often algorithmically curated. And what is served to users under the guise of choice frequently reflects the imperatives of engagement, profit, and emotional resonance over critical understanding or cognitive expansion.
The transformative potential of AI in democratising access to information comes with risks. Algorithmic enclosure and content curation can deepen epistemic inequality, particularly for the youth, whose digital fluency often masks a lack of epistemic literacy. What we need is algorithmic transparency, civic education in media literacy, and inclusive knowledge formats.
AI has made an undeniably positive contribution to the information ecosystem. Algorithmic tools translate obscure languages, summarise complex scholarly texts, and connect users with diverse intellectual and political content. Generative AI reduces barriers to comprehension and enables new forms of expression, particularly for users historically excluded from elite institutions or formal epistemic spaces.
Social media platforms have become potent sites of public and political education, disseminating alternative knowledges through short-form video reels, visual storytelling, and memetic discourse
Social media platforms — driven by sophisticated AI models — have become informal yet potent sites of public and political education. Movements, critiques, and alternative knowledges are disseminated not through peer-reviewed journals or political pamphlets, but through short-form video reels, visual storytelling, and memetic discourse. This has expanded the terrain of civic pedagogy, allowing people to engage with ideas from Marx, Foucault, Fanon, or bell hooks outside traditional academic settings. It also fosters transnational solidarities and offers affective anchorage in otherwise alienating social landscapes.
Yet 'access' in AI-mediated environments is rarely neutral, and quite unnatural. Recommendation engines, designed to maximise engagement, deliver content based not on civic or pedagogical value, but according to behavioural predictions. These systems learn from users’ micro-actions — likes, pauses, shares, scroll patterns — to curate a personalised, but often self-reinforcing, feed.
This process introduces epistemic enclosure. Users feel as if they are exploring the world, but they are circling within a curated informational ecosystem. A user drawn initially to content on mental health might find their feed increasingly filled with wellness trends, commercialised self-help advice, and productivity mantras rather than content that interrogates the structural determinants of distress. Similarly, a curiosity about religion in global politics might be redirected into polarising snippets far removed from historical analysis.
By privileging content that is easily digestible, AI systems encourage knowledge consumption that favours immediacy and novelty
The algorithm’s logic may not be intrinsically ideological, but its effect is political. By privileging content that is sensational and easily digestible, AI systems encourage knowledge consumption that favours immediacy over reflection, and novelty over depth.
The opacity of these systems compounds the challenge. Unlike traditional gatekeepers, algorithms do not name their principles nor explain their selections. Their authority is invisible and unaccountable, even as they become central to the formation of political subjectivities.
I am not a digital pessimist who advocates for the rejection of AI. But I do demand a more critical, transparent, and democratically accountable AI information architecture — one that acknowledges the epistemic stakes of algorithmic design. To make this happen, I suggest several policy priorities:
First, we must institutionalise algorithmic literacy as a foundational component of civic education. Citizens — particularly youth — should learn how algorithmic systems curate content. They must understand the mechanics of engagement metrics — such as clicks, likes, shares, and watch time — that influence content visibility and circulation. Cultivating algorithmic literacy equips individuals with critical competencies to interrogate these processes and intervene in their digital environments.
Second, social media platforms must meet higher standards of transparency and public accountability. As the digital sphere becomes a de facto public sphere, we must scrutinise its underlying architecture. Disclosing algorithmic design parameters, ranking criteria, personalisation weights, and optimisation objectives is essential to understanding how platform logics influence visibility hierarchies and shape public discourse.
We need a more critical, transparent, and democratically accountable AI information architecture which acknowledges the epistemic stakes of algorithmic design
Third, scholars and educators must reimagine knowledge production and sharing. The traditional journal article remains vital but is insufficient. Researchers should explore newer formats — visual essays, explainers, annotated videos — that bridge availability and accessibility without sacrificing rigour.
Fourth, generative AI tools like ChatGPT have transformed academic work. Rather than preventing students from using these tools, educators should raise the bar for human contribution. Assignments could ask students to critique the epistemic assumptions and biases embedded in AI-generated text, or expand upon generative outputs by incorporating empirical data, theoretical nuance, or counterarguments. The goal is to foster critical engagement, not merely detect 'cheating'.
Finally, we must cultivate a politics of attention that resists the commodification of attention and distraction. This involves creating digital spaces that reward slow, thoughtful engagement and nuanced dialogue.
Being shown the world is not the same as knowing it. AI has given us a sense of omniscience, yet social media platforms have often hollowed out the deeper practices of inquiry. We scroll endlessly, absorb selectively, and reflect intermittently. The world may be at our fingertips, but it remains out of reach.
The central task ahead is to ensure that information availability does not become a substitute for meaningful knowledge access. Political science and public pedagogy have critical roles to play — not merely in critiquing the algorithmic present, but in working with it to reimagine a more just and accessible epistemic future.
The author thanks Dipjyoti Paul for his valuable insights in the preparation of this blog piece
