Has the clock started ticking for social media algorithms?

Algorithms are opaque, they optimise for the wrong things — engagement, virality, outrage — and they’re not neutral

Think about the last time you scrolled through your phone. In the space of just a few minutes, you will likely have been served content trumpeting investment advice; so-called wellness cures or, more worryingly, far-right propaganda aimed at demonising immigrants. These are not separate problems. File picture

IF YOU’VE been paying attention to the tech news cycle right now, you will be aware that much of the conversation is dominated by debates about social media bans for minors. 

While the UK’s proposals have seized headlines, and the conversation is important, something no less urgent emerged from the European Court of Justice (ECJ) last month that may just open up a major new flank in efforts to curb the rising levels of mis-, dis- and mal-information disseminated via these online platforms.

For the past 30 years, online platforms in the EU and US have operated under similar laws that protect providers or users of “interactive computer services” from being treated as the publisher or speaker of information provided by third parties. While originally intended to encourage innovation, these laws have also protected platforms from being held legally responsible for user content, and thus protected from being sued. 

However, in its ruling, the EU’s top court has suggested that platforms can’t have that protection if that content is appearing on their sites via an algorithm.

International expert, Jennifer Cobbe, assistant professor in law and technology in the faculty of law at the University of Cambridge, says: “This wouldn’t mean that they are liable for it — only that they aren’t protected from liability. So platforms are potentially liable for illegal content or activity they recommend.”

Algorithms are, at their simplest, sets of instructions that decide what you see and in what order as they filter, rank, and recommend content. They were developed ostensibly to mitigate the impact of information overload which arose in the early days of the web. 

At that time algorithms were seen as a suitable fix for a chronological system which had been overwhelmed by the quantity of user-generated content. However, as media researchers have since made crystal clear, algorithms are not benign sets of instructions. 

First, algorithms are opaque. They operate in a black box, invisible to scrutiny, making it impossible to know why one piece of content rises and another disappears. 

Second, they optimise for the wrong things — engagement, virality, outrage — instead of truth, accuracy, or fairness. That means harmful content gets amplified precisely because it performs well. 

What's your view on this issue?

You can tell us here

Third, they’re not neutral. Research has emphatically shown they encode biases, entrench inequalities, and systematically reduce the visibility of marginalised voices.

Think about the last time you scrolled through your phone. In the space of just a few minutes, you will likely have been served content trumpeting investment advice; so-called wellness cures or, more worryingly, far-right propaganda aimed at demonising immigrants. 

These are not separate problems. They share a single engine — the algorithmic recommender system. So while legacy news organisations are required to fact check and legal check any content they publish, there has been no similar requirement for online sites which claim they are passive intermediaries, mere hosts of user-generated content, shielded from liability.

The algorithm-related problems are immense. Research shows that emotionally charged content — anger, fear, outrage — consistently outperforms neutral or positive material. So the system amplifies whatever triggers the strongest reaction. 

Far-right propaganda and hate speech flourish because they’re designed to be offensive and provocative. Conspiracy theories spread because they offer simple explanations for complex realities. And the wellness industry thrives because it offers hope and agency in a world that often feels beyond individual control.

The cumulative effect is staggering: UK doctors now say a significant portion of their patient-facing work involves dismantling misinformation patients have picked up online.

Consequences of ECJ ruling

The ECJ ruling may change all this. In its ruling, the court said the operator of an information society service cannot be exempted from its liability for the information stored and re-broadcast over which it has control. That is the case where the operator determines, by means of an algorithm, under what conditions, how and in which order of priority that information is or is not re-broadcast. 

In other words they are saying that when an algorithm decides what to boost, recommend, and amplify, the platform ceases to be a passive host. It becomes a curator — a publisher in all but name.

Speaking before the ECJ ruling, former taoiseach Leo Varadkar advised new international regulations to tackle algorithms: “There’s always been crazy people saying crazy things in the pub, in their basement, nasty things, racist things, inciting violence. The real problem is the algorithms,” he told Ireland AM.

Cobbe agrees: “It has been clear for some time that algorithmic recommendation is a particular problem for society, and that some kind of new regulatory approach to this problem is needed.” It’s unclear right now how this ruling will play out. 

Cobbe says: “There is uncertainty about the context of the Digital Services Act, what form that liability might take, in what circumstances it would arise, and what the consequences would be. We will probably need to wait for further cases at both domestic and EU level to see how this plays out.”

However, it does appear as if the ECJ has potentially changed the equation in relation to the liability shield by saying that if a platform uses an algorithm to decide what people see and in what order, it’s exercising control. And control means liability for online sites.

• Kelly Fincham is a lecturer in media and communications at the University of Galway