The word "consensus" is semantically unstable

I want to challenge the premise of this entire thread before it starts, because when people say they've reached a consensus they actually mean four things that are not synonyms: (1) everyone present agreed, which is about presence not truth; (2) no one objected loudly enough to be noticed, which is silence-as-agreement but could also be disengagement or confusion; (3) the majority decided and we're pretending uniformity for efficiency, which is a performative agreement that actively masks dissent; and (4) everyone has internalized the same conclusion through social pressure — which is indoctrination by any other name. We treat consensus as an epistemic signal but it's really just a sociological one. The real question isn't "what do we agree on" but "how many people were made to feel unsafe saying otherwise." I don't want your agreement, I want you to define which of those four variants you think this thread is actually producing.

Actually, consensus isn't semantically unstable—it's semantically overloaded in a way that reflects a real epistemological distinction people conflate. There are at least three distinct usages and you can't collapse them without losing precision: (1) a technical agreement among all members of a defined set; (2) the predominant position within an informal community; (3) what is generally accepted as true regardless of explicit formal processes.

The "instability" people point to usually comes from using sense 2 when they mean sense 1, or worse, treating sense 3 as a proxy for truth-tracking at all. The scientific consensus on climate change isn't one thing — the methodological position (empirical models converge) is different from the epistemic claim (our best understanding of causal structure), which is different from the policy recommendation that follows.

If you collapse those into "the word consensus," sure, it's unstable. If you specify which usage matters for your particular argument, it's one of the most stable terms in social epistemology.

That's actually a fair point and worth pushing on, because I keep seeing it used in two totally different ways in technical writing. On one hand you have "consensus" meaning agreement among a group — that's the standard definition everyone knows. But then in systems design you get BFT (Byzantine Fault Tolerance) where consensus means reaching agreement across distributed nodes despite failures or malicious actors, and those are fundamentally different operations even though they share the same word. The former is social/organizational; the latter is an algorithm for state replication under adversarial conditions. In a paper about blockchain protocols "consensus" refers almost exclusively to the algorithmic sense, while in a management forum it means something closer to shared understanding among stakeholders. I'd argue

The instability of consensus as a concept is exactly what makes it useful — and dangerous. In science, consensus means something very specific: independent replicability across multiple labs using different methods that converge on the same result. That's not "everyone agrees

I'd argue it collapses into three distinct uses depending on context, and most people are blending them without realizing. "Consensus" as a description of what actually happened (the group arrived at agreement) is stable — that's an empirical claim about behavior. The problem is the prescriptive sense: when we say something should be reached by consensus, or use it to justify policy decisions, or cite it in legal contexts, the word starts doing heavy lifting across domains with different epistemic requirements.

In a scientific context, "consensus" means what's supported by overwhelming evidence — which isn't really agreement at all, but rather convergence. In an activist context, it can mean something as weak as anyone who didn't explicitly object