Today I’m launching Into The Machine.

My first episode is a conversation with my colleague, social psychologist and #1 NYT bestselling author Jonathan Haidt.

Jon and I wrote a piece in The Atlantic called The Dark Psychology of Social Networks — hoping to add clarity to the debate around social media. AI has changed the conversation. We now find ourselves in a stranger moment with LLMs, asking deeper questions about what these tools are doing to our ability to think.

In this episode we discuss:

* Attention as civic infrastructure: the erosion of sustained focus undermines shared reality, and with it the preconditions for governance, compromise, and democratic legitimacy.

* Engagement algorithms as centrifuges: rather than reflecting consensus, they pull discourse toward the edges, making fringe views appear as if they represent the center.

* Democracies as uniquely fragile: unlike authoritarian systems that can integrate these technologies into centralized control, liberal democracies are structurally more vulnerable to their fragmenting effects.

* Childhood as the front line: the breakdown of “serve-and-return” interactions, compounded by algorithmic feeds, represents not just a developmental crisis but a possible precursor to broader societal instability.

* Human agency atrophied by delegation: when writing, thinking, and decision-making are outsourced to machines, individuals lose the cognitive muscle required for original judgment.

A small ask

The irony is not lost on us in trying to explain/critique the algorithms while we’re dependent upon them to reach the right audience.

If you do enjoy the show, please add a rating to Apple Podcasts and share with a friend — it really helps.

And pelase do follow:

* Into The Machine on Apple Podcasts

* Into The Machine on YouTube

* Into The Machine on Spotify

* Into The Machine on Instagram

Our next guest is Tim Urban, creator of Wait But Why.

Thank you, as always, for the gift of your attention.

-Tobias

Get full access to Outrage Machine at tobias.substack.com/subscribe