How the Ellsberg Paradox and Unknown Unkowns Apply to the Supreme Court's Stay Grant in NetChoice v. Paxton (TechLords v. Texas)
The Supreme Court just this morning granted the Tech Lords’ request to vacate the 5th Circuit’s stay of the district court’s preliminary injunction n NetChoice v. Paxton. The effect of this is to suspend the Texas law requiring Twitter and suchlike big companies to not ban people for political reasons. See https://www.justsecurity.org/wp-content/uploads/2022/05/21A720-Order.pdf for the grant by a 5-judge majority, Alito, Thomas, Kagan, and Gorsuch dissenting. Only Alito, joined by Thomas and Gorsuch, wrote an opinion.
A bit more detail on the procedural history. In 2021, Texas passed a law saying that the giant Internet platforms like Twitter and You-Tube could not kick people off for their political opinions. (The law had various related rules too, e.g. revealing the post-priority procedures Twitter uses.) The Tech Lords have joined together in a consortium called NetChoice to fight this kind of law. Paxton is the Texas Attorney-General, so when the Tech Lords sued to suspend the law, the case was called NetChoice v. Paxton.
NetChoice won a preliminary injunction from the one judge in the federal district court, meaning that the law was suspended until a full hearing would make the injunction permanent or lifted— but probably Texas would just give up and not bother with that. Texas then appealed to the three-judge 5th Circuit Court of Appeals. They have not issued their opinion, but in May 2022 they granted a Stay of the preliminary injunction, meaning that the Texas law is back in force. NetChoice then asked the 9-man Supreme Court to lift the stay, so that the old preliminary injunction would stop Texas from enforcing the law until the Supreme Court decided either not to take the case or took it up and issued a decision adverse to NetChoice. June 1, today, the Supreme Court granted the request, meaning the Texas law is again suspended.
This Substack post is about one of the legal issues involved, a technical one but easy to understand as an issue, though harder to know how to decide. One of the four standard criteria for getting emergency relief such as a Preliminary Injunction or a Stay is having a substantial likelihood of success on the merits—- that is being likely to win at the end of the day once the Court has read everything, held hearings, and so forth. (The other criteria have to do with who loses out in the meantime, and how much.)
But what does “substantial likelihood” mean? Let’s suppose we think it means a 40% chance of success (picking 40% just to be concrete, replace it with 20%, 60% or 80% if you like). Judge Alito would add some more words to that, and say “substantial likelihood of success under existing law”. Another way to put that is “likely to win just on the basis of rules and precedents we already have, without winning by breaking old precedents or setting new ones where there simply aren’t any existing precedents”.
The Ellsberg Paradox of economics enters in here. Before we get to the Paradox itself, let’s make this even more concrete. Suppose everybody thinks the 1st Amendment applies to case, and if it does, there is a 40% chance that it applies in such a way that NetChoice wins. This is one way to get a 40% probability of winning.
But suppose there is a 50% chance that the Court thinks the 1st Amendment applies, and there is an 80% chance NetChoice will win IF it applies,; but there is also a 50% chance the Court will think that the 1st Amendment DOES NOT apply, in which case there is a 0% chance NetChoice will win. Overall, the 50% chance of an 80% probability of winning with the 50% chance of a 0% probability of winning gives us a 40% probability of winning— .50(.80) + .50(0) = .40.
Should the Court treat both of these 40% probabilies the same for whether or not to grant emergency relief?
Back around 1960, Daniel Ellsberg, later famous for leaking the Pentagon Papers (as well as for having his psychiatrist’s office bugged by Nixon’s henchmen and for engaging in orgies) did a simple experiment on this topic. Though simple, it actually combined two psychological elements, Compound-Probability Aversion and Ambiguity Aversion, so I’ll separate them and just start with Compound Probability Aversion. Suppose we ask people in an experiment whether they would prefer a bet that gave them a 40% chance of winning a dollar (and 60% chance of getting nothing); or a 50% chance of winning a dollar with 80% probability and a 50% chance of winning nothing. More people will prefer the first, simpler, bet. They don’t like the “compound probability”. But the overall odds of winning are identical. See Wikipedia or Ellsberg, Daniel (1961). "Risk, Ambiguity, and the Savage Axioms" , Quarterly Journal of Economics 75 (4): 643–669. doi:10.2307/1884324. For a study which did separate out Compount-Probability Aversion, see Y. Halevy. “Ellsberg revisited: An experimental Study,” Econometrica, 75:503-536, 2007.
Justice Alito’s suggestion is also related to a curious idea of economist Frank Knight from back around 1940 called “Knightian uncertainty”. The terminology he used is highly misleading, because his two terms have different meaning in (a) How he uses them, (b) How modern economists use them, and ( c) How everyday language uses them (“risk” and “uncertainty”). So I will call the idea “known unknowns” versus “unknown unknowns” (did Secretary of Defense Donald Rumsfeld invent that?). Preferring Known Unknowns to Unknown Unknowns is called “Ambiguity Aversion”.
Suppose I win a bet if
(a) The next coinflip will come out Heads, and
(b) Justice Roberts’s wife prefers chocolate ice cream to strawberry.
I don’t know whether either of these things are true. They are Unknowns.
I do know that there is a 50% chance of event (a), Heads. I have seen lots of coinflips, and we all agree that half of them are Heads (at least to the nearest percentage; maybe it’s closer to 50.0001%), even though actually the next flip is Heads with either 0% or 100% probability and we just don’t know which yet (a compound probability!). So event (a) is a Known Unknown.
I feel much less sure about event (b), Mrs. Roberts’s preferring chocolate ice cream. This, too, has either 0% or 100% probability, and I just don’t know which. It would be false to say “I have no idea which she likes best”. I am quite sure I wouldn’t say there there’s a 1% chance she prefers chocolate. Nor would I say 98%. Indeed, if pressed, I might settle on 50%. To make “I might settle” more precise, let’s be concrete. What it really means is that if you forced me to bet my $100 against your $100 on either her liking chocolate or her liking strawberry, my choice, it wouldn’t matter to me which of them I bet on, but if it was either my $100 on chocolate versus your $101 on strawberry or my $100 on strawberry versus your $100 on chocolate, I’d prefer to bet on chocolate. Having to put my money where my mouth is would pin down my exact probability. (Note that I’m not allowed to say I don’t want to bet; I have to choose.) But I feel anxious because event (b) is an Unknown Unknown. (I’m expanding Rumsfeld here. The original use is when I don’t even know what event (b) is, and maybe I don’t even know I have to succeed in another event beyond (a) to win so I’m disappointed when the coinflip comes out Heads and I still lose. Here, the Unknown Unknown means I don’t know which way the event goes, but I do at least know I have to pass (b) to win my bet.)
Thus, I put 50% probability on the coin toss and 50% probability on chocolate, and I have an overall 25% chance of winning the bet (.25 = .50(.50)). But I feel uneasy about the chocolate bit.
Applied to a court decision, if the law is well established by numerous precedents, leaving only the evidence and special details of NetChoice v. Paxton, we have a Known Unknown. Looking at my fellow judges, and thinking about how I usually decide cases based on precedent, I think there is 100% chance the First Amendment applies, and a 40% chance the facts of this particular case will mean the requestor, NetChoice, will win.
But if the law is not well established, so we don’t know if the First Amendment will apply here, much less the particular details and how my fellow judges and I will feel about them, we have an Unknown Unknown as well as a Known Unknown. It might be like our earlier probabilities. Now suppose if the 1st Amendment applies, there is an 80% chance NetChoice will win, but if it doesn’t, there is 0% chance. Now I have to make a guess as to the probability my fellow judges, and myself too, will settle on a new precedent for how the 1st Amendment applies to this kind of case. I might settle on 50-50. In that case, as before, the compound probability, my best overall guess, is that Netchoice will win with probability 40% once everything is figured in. But I am uneasy about that. Tomorrow I might think it’s 70-30, just because I’m less sleepy or I hear a new argument I’ve never heard before.
So Justice Alito is saying that compound probabilities and Unknown Unknowns should make a difference in deciding whether to grant emergency relief. If
(i) the law is clear and there’s a 40% chance NetChoice will win,
NetChoice is more entitled to the Stay than if
(ii) the law is unclear, and as a result of guessing at the probability of what new precedent will come out and combining that with the probability NetChoice will win under the new precedent based on its particular details we get an overall compound probability of 40%.
So these ideas from economics are a way to make precise the issue here. Of course, that doesn’t decide the issue. We’d want to look at how the common law has treated the “substantial likelihood” criterion. And we’d want to know whether it’s good policy or not to make (i) different from (ii).
[This is where I reached by 7:50pm Mountain Time on June 1.]
NOTES: Justice Alito writes:
To be entitled to vacatur of the stay, applicants must show, among other things, a “substantial likelihood of success on the merits.” Alabama Assn. of Realtors v. Department of Health and Human Servs., 594 U. S. ___, ___ (2021) (per curiam) (slip op., at 5). Members of this Court have argued that a determination regarding an applicant’s likelihood of success must be made under “existing law,” Merrill v. Milligan, 595 U. S. ___, ___ (2022) (ROBERTS, C. J., dissenting) (slip op., at 1); Wisconsin Legislature v. WiWissconsin Elections Commission, 595 U. S. ___, ___ (2022) (SOTOMAYOR, J., dissenting) (slip op., at 1) (“existing precedent”)
Sotomayor is mis
Sotomayr is miscited. She said *summary reversal* of a state court required existing law. R
MILLIGAN:
Justice Kagan called the majority’s decision to stay an Alabama redistricting order “one more in a disconcertingly long line of cases in which this Court uses its shadow docket to signal or make changes in the law, without anything approaching full briefing and argument.”45 And Chief Justice Roberts refused to join the majority for the same reason. Though he expressed disapproval of existing precedent, he thought that should be resolved on the merits, not previewed on the shadow docket
“Shadow Docket Experiments,” Edward L. Pickup & Hannah L. Templin Notre Dame Law Review Reflections 32 pages Posted: 10 Feb 2022 Last revised: 4 Mar 2022.GOOD.