Профиль

Interesting! But seems the doi URL isn't working yet. Could you share a direct link?

That paper argues that the literature on 'misinformation' has been studying what is in fact propaganda. These two papers are very much singing the same song (Also, if I were to call them "radicalization is radicalization" and "propaganda is propaganda", the intended audience would not be reached :)

Concepts are never politically innocent. We must call things what they are. It's not polarization; it's the radicalization of the political right. Open Access: doi.org/10.1017/S153... @poppublicsphere.bsky.social

Then meeting them halfway is not a remedy. It can be complicity. Then treating both sides as equivalent is not neutrality. It can be normalization. Then conflict and resistance are not the disease. They can be part of the solution.

Why does it matter what term we use? Because diagnoses imply remedies. If the problem is polarization, the solution is civility. Dialogue. Lowering the temperature. Finding common ground. Compromise. But what if the problem is right-wing radicalization?

“Radicalization” offers a better framework. It is directional and asymmetric. It asks how identities harden. How leaders construct existential threats. How norm-breaking spreads. How groups escalate. How unthinkable ideas enter the mainstream. It describes the process we are actually seeing.

Finally: ⬅️➡️ mutual repulsion. Polarization implies that the two sides push each other apart. But again, the evidence often shows the opposite. The rise of the radical right has produced a gravitational pull. Mainstream parties are pulled toward it, adopting its issues, frames and positions.

Among voters, the central dynamics are not about shifting opinions. They are about identities hardening. Loyalty to a leader. A sense of threat. A division between "the real people" and the enemies said to threaten them. That is not merely movement along an ideological line.

Among right-wing elites, the rules of the game are now at stake. Rejecting election results, capturing courts and media, and abandoning democratic restraint are not "more extreme positions." The shift is from competing within democracy to seeking to dismantle it. Distance is the wrong metaphor.

Second: 📍↔️📍 space Polarization imagines politics as positions on a line. People and parties move left or right, and polarization grows as the distance between their positions increases. But today's political transformation is not a divergence along a shared ideological scale.

One side trying to dismantle democratic institutions, and the other trying to stop them, are not symmetrical political movements. Calling this "asymmetric polarization" does not solve the problem. A concept built around two poles has to be stretched until it means its opposite.

But consider the first assumption: 🔵🔴 symmetry. Empirical research shows that the crisis has been driven overwhelmingly by shifts on the political right: authoritarianism, election denial, conspiracy theories, norm-breaking and tolerance of political violence. That is not two poles moving apart.

The appeal of the concept for academics is easy to understand. It offers the appearance of neutrality. Equal responsibility. Like journalism’s "both sides" logic, it lets us diagnose political crisis while appearing detached, balanced and scientific.

The concept of 'polarization' comes from physics. 🔵🔴 It describes two symmetric poles. 📍↔️📍 Moving apart in space. ⬅️➡️ Mutually pushing each other in opposite directions. This metaphor has come to shape how we understand democratic crisis. But every part of it is wrong.

Interesting! Thanks for sharing! Will check with this in the revision, e.g. incl. smaller models. I'm working on a paper looking at the temporal evolution of sycophancy - and you're right that it's becoming more prominent. I would assume that it's related to e.g. RLHF rather than just size though?

The bias literature has assumed that bias is a fixed political position that we can put a number on. But if the model action is context dependent, that becomes ill-defined. Ultimately, a useful measure would need to examine how the models behave in real-world conversations. That's hard, but doable.

The previous literature says that LLMs are left-wing. This paper shows that this might be an effect of them thinking that academics are left-wing, and adapting accordingly. The main takeaway is that we've been measuring "bias" poorly.

Hard to say anything definite unfortunately. The most immediate explanation would be what you suggest: academics are generally described in public discourse as being left-wing.

Yes, that’s the main argument of the paper. But it’s also hard to see a solution.

Check out the preprint: arxiv.org/pdf/2604.27633 We welcome your comments and feedback!