The article I want to write right now would talk about the fact that the scale of 2020 election misinformation was enormous, but claims, frames, and players were predictable down to absurdly specific levels. And that means we might be looking at detection and enforcement wrong.

I maybe need the help of some policy people on this, but basically you have a couple ways of going at things in info communities.

One is very general -- we look for these attributes and do our calculation. Harm level 6, deception level 9. Looks bad. Harm level 2, deception level 4: passable.

These sorts of general rules are what you're stuck with if you're dealing with endlessly novel exploits.

But what if you're not dealing with that in a given domain? What if you know there's going to be a story about someone throwing away ballots that is actually a joke but will be amplified as real, for example? What if that happens every election?

What if every election there will be claims that machine totals were hacked? If every election will feature a video of someone stuffing multiple ballots into a ballot box? Of decontextualized altercations with poll workers?

In these cases there are still competing goods here -- the allegations may be true, ballots may have really been stuffed, altercations may show illegality. But it occurs to me that if we know the specific stuff that is going to be shared we can produce specific *guidance*.

Not, "what level of harm does this video present" but rather "Under what circumstances do these poll worker altercation videos cross a line?" And also: "When a poll worker altercation video surfaces, what's the verification process?"

Similarly, with detection, we can supplement a lot of fancier scans looking for the novel stuff with scans for the stuff we know will come up, from poll worker altercations to the inevitable "Ha, ha, I'm a poll worker throwing away votes" joke that gets read as real.

Similarly, if the players are the same every time, maybe we need to focus much more on repeat offenders and much less on one-time offenders.

In short, we know a lot in advance about the who, what, when, where, and how of misinfo when it comes to electoral processes -- so more narrow and specific measures seem to be a better fit for these scenarios.