One of the persistent risks in any model of digital democracy is that people drift into ideological dead ends. The pattern is well documented. Someone encounters content that confirms their worst suspicions about the world. The algorithm serves them more of the same. Gradually the radical starts to feel like the only honest position, and the bridge back to the wider community quietly disappears.
It would be dishonest to pretend that direct digital democracy is immune to this. Any platform that opens up civic participation also opens up the possibility of capture by motivated fringes, coordinated bad actors, or communities that have retreated so far into their own certainties that engagement with others feels like betrayal.
The conventional response is to block, ban, and shame. It does not work. People who feel excluded from a platform do not moderate their views - they find somewhere else and become more convinced than ever that the mainstream cannot be trusted.
Audrey Tang's work in Taiwan offers a better starting point. The tools she helped build - particularly Pol.is, a platform for mapping opinion across large groups - were designed to surface unexpected areas of agreement rather than to amplify division. When you show people that someone they assumed was their opposite actually shares several of their priorities, the conversation changes. The bridge back becomes visible.
But bridging is only half the problem. We also need what might be called rituals of return - structured ways for people who have gone out to the edges to come back in, bringing what they have found, and having it genuinely tested rather than simply dismissed.
What might that look like in practice? Randomly selected deliberative groups that ensure radical positions are exposed to a representative mix of people, so no one stays permanently inside a monoculture. Formal dissent mechanisms that channel outlier views into reports that are interrogated and fed back into the main stream of debate, rather than quarantined. Reputation systems that reward constructive contribution across multiple topics, so trust built in one area carries weight in another. Transparent ways back - the right to publish a minority report, propose a policy pilot, have an argument stress-tested rather than silenced.
The point is not to sand politics down into bland agreement. Disagreement is the raw material of democratic life. The question is whether the system gives disagreement somewhere productive to go, or leaves it to fester in isolation until it has nowhere to go but outward.
Direct digital democracy will only be genuinely robust if it treats radicalisation not as a pathology to be purged but as a stress test - an opportunity to strengthen the bridges and ensure there is always a route back into shared civic life.