My talk at the RESAID event: “Information Disorders in the Age of AI and Algorithms”

A few days ago, I was invited to serve as a panelist at a RESAID project (“Creating Societal Cognitive Resilience Against Information Disorders – RESAID”) event. This EU-funded project has substantive online course content and has developed a few online games related to disinformation issues. The panel was reported by AA as well.

Here are my highlights from my talk:

Information Disorders in the Age of AI and Algorithms

We are living through a profound transformation in how information is produced, circulated, and authenticated. The shift from a traditional media ecosystem—dominated by editors, institutional gatekeepers, and professional norms—to an algorithmic environment driven by engagement metrics has altered not only our media habits but also our collective sense of truth.

What follows is a short reflection based on my recent panel talk on information disorders in the age of AI and algorithmic governance.

From Gatekeepers to Algorithmic Selectors

In the traditional media era, editors and newsrooms played the central role in shaping public information flows. Publishing costs and institutional ethics created a natural barrier that limited the spread of misinformation.

Today, algorithms have replaced editors as gatekeepers. They operate without transparency, and their logic prioritizes engagement over accuracy. Virality wins over verification. Each user now inhabits a personalized information bubble—leading to the erosion of a shared reality.

Content Moderation at Scale: A Structural Challenge

Platforms struggle to control harmful content, but several structural issues make true moderation nearly impossible:

• Scale: Billions of posts circulate daily.

• Context: Algorithms misread irony, satire, slang, and cultural nuance.

• Inconsistency: Rules differ across platforms and countries.

• Opacity: Decisions are rarely explained; researchers lack access.

As Tarleton Gillespie notes, content moderation at scale is inherently unsolvable. Automation introduces new problems rather than eliminating old ones, especially in non-English contexts.

The New Risks Introduced by Generative AI

Generative AI tools—LLMs, image generators, voice synthesis—magnify existing information disorders:

• Deepfakes undermine trust in audiovisual evidence.

• LLM hallucinations threaten reliability in scientific and factual domains.

• AI becomes an epistemic mediator, influencing what knowledge we access and how we interpret it.

This marks a shift from misinformation being solely a human phenomenon to becoming a machine-mediated one.

AI Grooming: A New Frontier in Manipulation

One emerging threat is AI grooming—deliberate attempts to manipulate AI models by poisoning or “seeding” the datasets they rely on.

Instead of targeting people directly, actors target the models that people use.

This includes:

• Flooding the web with political narratives so LLMs reflect them.

• Strategically influencing datasets to shift model behavior over time.

• Attempts by state and non-state actors to bias AI systems to shape public opinion indirectly.

In other words:
manipulators groom the machines so that the machines, in turn, groom the public.

This is subtle, infrastructural, and extremely hard to detect.

Societal Implications: Fragmented Publics and Eroding Trust

These shifts create deep societal consequences:

• Declining trust in institutions

• Fragmented and polarized public spheres

• Heightened vulnerability during election periods

• Increased uncertainty during crises (e.g., earthquakes, political turmoil)

Truth now competes with engagement-optimized falsehoods on unequal ground

What Can Be Done?

There is no simple solution, but several measures can improve resilience:

• Rethink fact-checking to match the speed and scale of digital misinformation.

• Demand greater algorithmic transparency.

• Expand media and AI literacy, including platform logic and model behavior.

• Enable independent research into platform data and moderation systems.

These steps won’t eliminate information disorders but can reduce their impact.

Conclusion: A New Epistemic Regime

Information is no longer produced only by humans. AI systems now participate in generating, interpreting, and legitimizing knowledge. Understanding this shift is essential for rebuilding trust and designing democratic, accountable digital infrastructures.

Ultimately, the future of the public sphere depends on how we regulate, design, and critically engage with the AI and algorithmic systems that mediate information.