Synthetic Consensus

Definition of Synthetic Consensus

Synthetic consensus is a machine-produced group agreement: an AI system takes people's separately-stated views and computes a "group statement" optimized for the widest acceptance, rather than having them argue toward one. It replaces the friction of debate with algorithmic optimization, treating agreement as something to be calculated rather than reasoned toward.

This distinguishes it from autonomy, the idea that people should make their own choices. Synthetic consensus does not force anyone, but it weakens autonomy by taking away the chance to use it. When agreement is calculated, no one needs to think about why they believe what they do or defend it. It also differs from public participation, which means having people involved in decisions that impact them. Synthetic consensus can seem like participation since everyone’s input is gathered, but the value of participation comes from discussion. In this case, the inputs are handled privately and never interact. People share their views without truly debating together.

Implications of commitment to Synthetic Consensus

Committing to limit synthetic consensus means rejecting the idea of treating a widely accepted output as genuine collective reasoning. The key issue is the difference between the agreement people reach and the agreement determined for them. Only the first captures views tested against opposing reasons. The calculated version has a built-in bias toward the status quo. A system that optimizes for acceptability discards the least acceptable inputs. These often represent the minority viewpoints that can reshape a debate.

This leads to three design requirements. First, the tool must preserve the reasons behind positions, not reduce them to a midpoint. Second, it must clearly communicate its trade-offs, including whose views were compromised and why. Third, it must serve as a tool people can question, not as a final answer that provides a completed response.

Societal transformations required for addressing concern raised by Synthetic Consensus

The problem is more subtle. If we get a “consensus” in seconds from a machine, we will lose our skill in discussing things with each other. This skill becomes atrophied if you do not practice it. Gradually, we will forget how to be in a situation of disagreement, to hear another point of view, and reconsider it. This is not an automatic process; it is a skill which can atrophy if a machine works for you. There is another threat. If an artificial intelligence constantly gives us the “middle” decision, we begin to think that it is right simply because it has caused the least controversy. The most undisputed decision is not necessarily the right one. Societies sometimes believed in something that was false until one voice was heard.