Rebooting Democracy: How AI and Digital Platforms Can Move Us Beyond Likes

The Rectangle of Sadness

Most of us know a different version of digital engagement. Comedian Hasan Minhaj called it the “rectangle of sadness” — that glowing screen we stare at, absorbing national politics we feel powerless to influence.

Social media platforms weren’t designed for democracy. They were designed for ad revenue. The result is a civic experience that looks like participation but functions more like spectatorship — clicking, liking, sharing, and feeling vaguely furious about things we can’t change.

Political theorist Benjamin Barber had a sharper phrase for it: “the caverns of private solitude.” Lots of noise. Very little power.

The Original Promise (And Why It Stalled)

Back in 1999, the internet felt genuinely revolutionary for democracy. Vint Cerf — one of the web’s architects, rarely spotted without his three-piece suit — noted with some delight that legislators had expected the internet to be another broadcast channel. Then they discovered constituents could talk back.

That two-way energy inspired real experiments. Working with Barber, Beth Simone Noveck built Unchat — the first software purpose-built for democratic deliberation. It wasn’t a comment section. It rotated moderation between participants on a timer, structuring fairness directly into the interface. The goal wasn’t talk for its own sake. It was collective decision-making.

It was a genuinely interesting design. And then the next two decades happened.

What Didn’t Change

Platforms multiplied. Reach expanded. Institutions stayed exactly the same.

Despite decades of civic tech experiments and an explosion of online engagement tools, two-thirds of people still believe they have little influence over government decisions. The infrastructure for participation exists. The institutional will to act on it largely doesn’t.

This is the gap that matters — not the technology, but the connection between digital input and real-world power. Brazil’s legislative petition system is rare precisely because it closes that loop. Most platforms don’t bother trying.

Where AI Changes the Equation

Here’s where this moment feels genuinely different from 1999.

AI doesn’t just help people communicate — it can help institutions process what citizens are actually saying. Summarizing thousands of public comments. Identifying patterns in constituent concerns. Translating policy complexity into plain language that invites real feedback rather than discouraging it.

The bottleneck in democratic participation has rarely been citizen willingness. It’s been institutional capacity to listen at scale. AI, applied thoughtfully, attacks that bottleneck directly.

The risk, of course, is the same as always: tools designed to simulate engagement rather than enable it. A chatbot that feels responsive but routes nothing to anyone with actual authority is just a more sophisticated rectangle of sadness.

The Design Question Nobody Asks Enough

Technology doesn’t save democracy. Design choices do.

Unchat’s rotating moderation wasn’t accidental — it was a deliberate structural decision to prevent any single voice from dominating. Brazil’s petition platform wasn’t just a form — it was wired into a legislative process that had to respond. These details are everything.

The question worth asking of any civic AI tool isn’t “does it engage users?” It’s “does it give citizens actual power to shape decisions?” Those are very different products. Most of what gets built answers the first question and ignores the second.

The Takeaway

Thirty years of online democracy experiments have taught one consistent lesson: the technology is rarely the hard part.

The hard part is building institutions willing to act on what they hear — and designing platforms that demand they do. AI gives us new leverage on that problem. Whether we use it to manufacture the feeling of participation or the reality of it is still very much an open question.

Joca’s Law passed because a platform was wired to a process that had teeth. That’s the model. Everything else is just engagement metrics.