Popular ≠ credible or true

Part of Australia's Digital Duty of Care Bill has been described as an ‘off switch for the algorithm’. What does that actually mean? An option to have a persistent following-only feed.

But the bill doesn't explicitly require that feed to be chronological. So despite the headlines, switching off recommendations doesn't necessarily switch off algorithmic ranking. It just changes what the algorithm has to choose from.

A following-only feed limits an important way we find new voices.

And discovery matters in a fragmented media landscape where much of our news and information is consumed through platforms. Recommendations can expose us to perspectives we might not otherwise encounter, including ones we disagree with.

A following-only feed isn't necessarily an escape from echo chambers, either.

We tend to seek out people who share our existing beliefs. Some research suggests algorithmic recommendations can create networks that are less homogeneous than those formed through existing social connections.

Giving people more control over their feeds is a good thing. But I'm not convinced switching off recommendations improves our information environment.

The bigger question is how these systems decide what deserves our attention. And as AI increasingly helps us find and interpret information, how do we ensure it surfaces credible sources rather than simply popular or engaging content?

Popular ≠ credible or true.

Perhaps the better question is what we want these systems to optimise for in the first place.