"There are no inconsistencies across adjudicators because there’s one adjudicator."
This seems like a bold claim, when AI can be inconsistent with itself!
Given a complex prompt and difficult questions, which this seems like, it's totally possible that an AI would come down on one side XX% of the time and YY% on the other.
Sure you could go one level deeper and have AI adjudicate 10 times and take the mode, or flag cases that are inconsistent for manual review, but I think assuming AI totally eliminates this problem is simply incorrect.
"A ruling for the veterans will significantly increase the number of veterans who get benefits, which will on average increase the accuracy of benefits adjudications, because many veterans are currently being short-changed."
As of 2021, nearly 20% of veterans were receiving disability compensation payments.
"Likewise, there’s no such thing as de novo harmless error review. The appellate court isn’t reviewing anything when it conducts a harmless error analysis. The trial court wouldn’t have conducted a harmless-error analysis because, presumably, it didn’t realize it was making an error."
In this context, the "trial" court is reviewing an administrative decision that the plaintiff must show was prejudicial in order to prevail; if the trial court determines that any error in the administrative determination was harmless, that harmless-error analysis would be subject to de novo review, would it not?
I think here the "trial" court is the Board and the "appellate" court is the Veterans Court. The Federal Circuit - which hears appeals of Veterans Court decisions - only hears issues of law, I don't think it would hear a challenge to the Veterans Court's harmless-error assessment.
Thanks for explaining this. I love learning about the legal questions buried in these more opaque cases. It's such a good explanation for non-lawyers.
One question I have -- why was this granted cert in the first place? There's no circuit split, it's only ever going to affect the Federal Circuit. It's hardly the most pressing question before the court and it's a close call, as you say. The Federal Circuit's decision seems perfectly reasonable, so why not let sleeping dogs lie?
The Federal Circuit has exclusive jurisdiction over the subject matter within its ambit, so waiting for a circuit split is not a useful heuristic for thinking about when cert should be granted on any of their decisions.
I don't have a great sense for what the Court is looking for when deciding whether to grant cert on Federal Circuit appeals (they basically never take any involving, say, government contracts cases, for example), but I don't think it's a bad thing for them to be a bit more liberal towards taking up appeals out of that particular jurisdiction given the absence of other correctives.
I'm unsettled by "the anticipated practical consequences of the Court’s ruling" should it decide that Congress passed a law that does nothing. I'm more inclined toward that reading of the statute, but I think the rule against surplusage has underrated beneficial implications for legislative drafting and good governance. The judiciary should be exceptionally wary of (1) allowing Congress to pass laws that have no practical effect and (2) giving lower courts a tool for permitting policy considerations to suffuse statutory interpretation by finding an absence of meaning, where convenient.
Interesting but doesn't seem to get to real question, which is who is footing the legal costs for these lawsuits and why. There doesn't seem to be any "injustice" going on here - - the fact patterns suggest that both these individuals were fairly treated and equitably compensated regardless of how you read the law. In fact, the mere fact that these are the fact patterns being litigated seems to suggest that the system, cumbersome though it may be, works quite well. So, again, from a practical perspective, the real question is why someone is willing to spend their time on this legal question and whether only reason SCOTUS is looking at this is to show deference to veterans.
One underexplored theme from this is AI as getting around the disfunction of government employment contracts: we keep the people, and they still get to shuffle papers for a while, but their actual responsibilities are delegated to AIs, and as they retire they aren't replaced.
Maybe won't be just governments doing that, although I think there is a particular use case given than governments basically never sack people.
Excellent and thought- provoking article. Makes it seem like, so far, the LLMs have a bias against predictive humility and for "always provide an answer, even if it's BS, but don't provide a disclaimer that it's likely BS."
"There are no inconsistencies across adjudicators because there’s one adjudicator."
This seems like a bold claim, when AI can be inconsistent with itself!
Given a complex prompt and difficult questions, which this seems like, it's totally possible that an AI would come down on one side XX% of the time and YY% on the other.
Sure you could go one level deeper and have AI adjudicate 10 times and take the mode, or flag cases that are inconsistent for manual review, but I think assuming AI totally eliminates this problem is simply incorrect.
Learn about the legal deception history in my podcast here:
https://soberchristiangentlemanpodcast.substack.com/p/s2-ep-6-legal-deception-the-magic
"A ruling for the veterans will significantly increase the number of veterans who get benefits, which will on average increase the accuracy of benefits adjudications, because many veterans are currently being short-changed."
As of 2021, nearly 20% of veterans were receiving disability compensation payments.
https://www.census.gov/newsroom/press-releases/2021/veterans-report.html
I would guess a large share of successful claims of disability are bullshit. What makes you think veterans are being short-changed?
I have no idea! There’s no way to test these hypotheses, so the Court shouldn’t even try.
"Likewise, there’s no such thing as de novo harmless error review. The appellate court isn’t reviewing anything when it conducts a harmless error analysis. The trial court wouldn’t have conducted a harmless-error analysis because, presumably, it didn’t realize it was making an error."
In this context, the "trial" court is reviewing an administrative decision that the plaintiff must show was prejudicial in order to prevail; if the trial court determines that any error in the administrative determination was harmless, that harmless-error analysis would be subject to de novo review, would it not?
I think here the "trial" court is the Board and the "appellate" court is the Veterans Court. The Federal Circuit - which hears appeals of Veterans Court decisions - only hears issues of law, I don't think it would hear a challenge to the Veterans Court's harmless-error assessment.
Thanks for explaining this. I love learning about the legal questions buried in these more opaque cases. It's such a good explanation for non-lawyers.
One question I have -- why was this granted cert in the first place? There's no circuit split, it's only ever going to affect the Federal Circuit. It's hardly the most pressing question before the court and it's a close call, as you say. The Federal Circuit's decision seems perfectly reasonable, so why not let sleeping dogs lie?
The Federal Circuit has exclusive jurisdiction over the subject matter within its ambit, so waiting for a circuit split is not a useful heuristic for thinking about when cert should be granted on any of their decisions.
I don't have a great sense for what the Court is looking for when deciding whether to grant cert on Federal Circuit appeals (they basically never take any involving, say, government contracts cases, for example), but I don't think it's a bad thing for them to be a bit more liberal towards taking up appeals out of that particular jurisdiction given the absence of other correctives.
I'm unsettled by "the anticipated practical consequences of the Court’s ruling" should it decide that Congress passed a law that does nothing. I'm more inclined toward that reading of the statute, but I think the rule against surplusage has underrated beneficial implications for legislative drafting and good governance. The judiciary should be exceptionally wary of (1) allowing Congress to pass laws that have no practical effect and (2) giving lower courts a tool for permitting policy considerations to suffuse statutory interpretation by finding an absence of meaning, where convenient.
On board with AI replacing the VA.
Interesting but doesn't seem to get to real question, which is who is footing the legal costs for these lawsuits and why. There doesn't seem to be any "injustice" going on here - - the fact patterns suggest that both these individuals were fairly treated and equitably compensated regardless of how you read the law. In fact, the mere fact that these are the fact patterns being litigated seems to suggest that the system, cumbersome though it may be, works quite well. So, again, from a practical perspective, the real question is why someone is willing to spend their time on this legal question and whether only reason SCOTUS is looking at this is to show deference to veterans.
Thanks for this!
One underexplored theme from this is AI as getting around the disfunction of government employment contracts: we keep the people, and they still get to shuffle papers for a while, but their actual responsibilities are delegated to AIs, and as they retire they aren't replaced.
Maybe won't be just governments doing that, although I think there is a particular use case given than governments basically never sack people.
Excellent and thought- provoking article. Makes it seem like, so far, the LLMs have a bias against predictive humility and for "always provide an answer, even if it's BS, but don't provide a disclaimer that it's likely BS."