The Appeal

Show notes

Every time a platform removes a post, suspends an account, or quietly demotes a piece of content, it has made a ruling. The language we use hides this: we say a system "enforced a policy" or "took an action," as if no judgment occurred. But to decide that one person's speech may stay and another's may not is to adjudicate — to apply a standard to a particular case and hand down a result that has winners and losers. The distinctive fact of our moment is that this adjudication happens at a scale no court has ever faced: not thousands of cases but billions, decided in milliseconds, largely by automated systems classifying content faster than any human could read it. The question of this episode is the oldest one in the law, asked of the newest kind of tribunal: when a system rules against you, what does it owe you?

The classical answer is due process, and its components are unglamorous and concrete. You are owed notice that a decision was made and what it concerned. You are owed a reason — not a slogan but a stated basis specific enough that you could understand why this case came out this way. You are owed an appeal to someone or something that did not make the first call. And underneath all of it you are owed a reviewable record: a trail showing what was decided, on what grounds, and against which rule, so that the decision can be examined rather than merely accepted. None of these requires that the ruling be correct. They require that it be the kind of ruling a person could engage with — that there be a "there" there to contest. Scale does not dissolve that obligation; it only makes meeting it harder, and makes the temptation to skip it stronger.

This is why "your content violated our guidelines" — with no reconstructable basis — fails the person it rules against, however true it may be. A conclusion offered without the path that produced it cannot be challenged, because there is nothing to take hold of: no statement of which rule, applied to which feature of the content, on what evidence. The affected user is left in the position of someone told only that they lost, never why or how, and asked to accept that the system that ruled is also the only authority on whether it ruled well. The standing problem is acute precisely because the cases are many and each is small: the very scale that makes individualized reasoning expensive is the scale that makes its absence feel, to any single user, like being processed rather than judged. The tension between handling everything and being contestable in any particular is the real subject here — not whether platforms should moderate, which they plainly must, but what a defensible moderation decision is made of.

This show's through-line is that a conclusion is legitimate not because it was reached but because the path to it can be reconstructed and contested — and content moderation is where that principle meets its hardest test of volume. The book this series accompanies, Admissible Reality, argues for systems whose outputs come with the warrant behind them: not just a verdict but the reviewable basis a careful objector could have attacked. A removal that arrives with notice, a specific reason, a route of appeal, and a record is a ruling someone has standing to contest; one that arrives as a bare result is an exercise of authority that has quietly declined to be accountable. The conversation is principled, not partisan — about process and standing, not about any particular platform, viewpoint, or controversy. It asks a trust & safety lead what it actually takes to give an automated adjudication the qualities that make a human one fair.