Understanding the Voting System

Most online feeds say they are powered by "voting." In practice, they are powered by reactions.

The interface asks a loose question—like, dislike, upvote, downvote. It records views, shares, and watch time. These signals are treated as if they were expressions of collective judgment, but they are not the same thing. A "like" can mean "I agree," "this amused me," "my friend posted this," or simply "I tapped by habit." A long watch time can signal fascination, outrage, or confusion. None of these are deliberate statements about what each community as a whole should be seeing.

The problem is not bad faith; it is ambiguity. When the question is unclear, the answers are noisy. When there is no cost to careless voting, there is little incentive to think beyond one's immediate reaction. And when engagement metrics are treated as votes, the system rewards whatever captures attention, regardless of whether it informs, clarifies, or represents the considered view of the group.

Our approach begins from a different premise. Even the simplest democracy has four elements: a clearly defined question, a defined set of voters, equal weight of votes, and an outcome determined by majority rule. Feeds rarely have any of these. We built ours to approximate them as closely as a real-time platform allows.

From Expression to Prediction: The Prediction-Based Rating System (PBRS)

At the core of Veridonia is what we call the Prediction-Based Rating System (PBRS). Each voting round asks a narrow and explicit question: will most people here think this was worth seeing? Nothing more. Not whether it is true in some abstract sense, not whether one personally agrees, but what the community would collectively say is worth its attention. The question is a prediction on purpose: rating moves according to how well a reviewer reads their community, so influence accrues to people who judge well on its behalf.

Ordinarily, a small group of members is randomly selected to answer that question. They do not choose the post; the system assigns it. After they vote, the majority outcome becomes the reference point. Individual ratings are then updated according to how well each participant's vote matched that outcome.

This turns voting into a form of forecasting. Each decision is implicitly a prediction about what the community, consulted more broadly, would decide. Participants who consistently anticipate that outcome gain rating. Those who consistently diverge lose rating. Over time, rating becomes a measure not of authority or virtue, but of predictive reliability.

The mechanism is deliberately relative and zero-sum. Rating cannot accumulate indefinitely. It must be maintained. Early arrival confers no permanent advantage; inactivity and misalignment gradually reduce standing. Influence flows toward those who demonstrate, repeatedly, that they understand how the community judges content, not those who shout the loudest or mobilise the most attention.

Rating affects influence in voting and posting, but not access to reading. The system shapes what rises to prominence; it does not restrict what individuals may view.

Approximating a Referendum: The Multi-Round Voting Process (MRVP)

If we wanted perfect democratic legitimacy, we would ask every member to vote on every post. That would also make the feed unusably slow. The Multi-Round Voting Process (MRVP) is our compromise: an approximation of a referendum, structured to remain practical at scale.

In a small community there is one round: a random panel's majority settles the post. Once a community is large enough to draw two panels, the decision is split. In the first round, a randomly selected group evaluates the post. The goal here is filtration. Clearly, low-signal material is removed early, with minimal cost to the broader community. Because selection is random, no one can reliably insert themselves into specific decisions, and coordinated manipulation becomes uncertain and expensive.

Posts that pass move to a second round. Here, a smaller group of higher-rated participants makes the final decision. Their role is to act as a statistical proxy, not a permanent elite. Because their ratings reflect a track record of alignment with community outcomes, using them reduces the number of required votes while preserving the expected result of a much larger consultation.

This second round is not fixed in composition. Ratings change continuously. Participants can rise or fall based on performance. The structure remains bottom-up: representativeness is earned and defended, not granted once and held indefinitely.

Together, PBRS and MRVP allow the platform to simulate the logic of a referendum—clear question, defined voters, majority rule—without asking everyone every time.

There is one exception, closer to a referendum. An editor may put their own post to everyone instead of a random panel: every member who can be reached when it is submitted is asked, in a single round, and a simple majority of the votes cast decides, whatever their number. A tie, or no votes at all, does not publish. It stays open for up to a day, or until everyone asked has answered, and standing moves exactly as it does after a panel's decision. Only an editor may ask, only about their own post, and at their own stake.

Why This Was Necessary

It makes the feed reflective of what the community, taken as a whole, is likely to consider worth attention — not what generates the strongest immediate reactions. This allows a single, diverse community to converge on shared standards, rather than fragmenting into separate spaces divided by narrow biases.

It also changes incentives. When influence depends on predicting collective judgment, participants are pushed to look beyond their own preferences. Modelling the broader community becomes more rewarding than signalling to one's immediate circle. Over time, this reduces the payoff of extreme or purely partisan content that cannot survive wider scrutiny.

Manipulation also becomes more costly. Random selection prevents actors from targeting specific posts at specific times. Because influence must be earned through accurate prediction over many rounds, creating disposable accounts yields little durable power.

And because a post that reaches the feed has ordinarily passed a randomly drawn panel's majority — and, in a large enough community, a second one — it cannot rely solely on energising a narrow slice of the community. To survive the full process, it must be understandable and acceptable to a broader cross-section of participants. In practice, this creates pressure toward material that is clearer, more proportionate, and less dependent on partisan reflex. Content capable of withstanding scrutiny beyond its immediate base.