Reply to Board of Veterans Appeals (Attorney Advisor)

Forum Index » Jobs and Careers

Reply to "Board of Veterans Appeals (Attorney Advisor)"

Subject:

Emoticons





More smilies

Text Color: Font: Close Marks

[quote=Anonymous][quote=Anonymous][quote=Anonymous][quote=Anonymous][quote=Anonymous][quote=Anonymous][quote=Anonymous][quote=Anonymous][quote=Anonymous][quote=Anonymous][quote=Anonymous]I think people here are catastrophizing. We don’t know what will happen, so I’m suspicious of anyone who announces they know the number with certainty. But we are in a better position than a lot of people. Clearly we are vulnerable to the RIF, some will probably lose jobs, but I do not think we are the priority for deep cuts. Yes, a lot of agencies have cut past the point of functioning, but VA is cutting less than other agencies percentage wise. That tells me this administration cares about our mission enough to not want to destroy it. We are mission critical. The project 2025 conservative agenda for VA is to privatize VHA and automate VBA. We don’t feature. So my guess is the RIFs focus on VHA to break it to the point of facilitating privatization. I think there will likely be a push to utilize AI more at VBA with an eye towards automation, but I don’t think it will get far because AI hallucinates. I don’t think they’ll make moves to automate our jobs unless it works at VBA and there’s no big pushback.

My most hopeful prediction is maybe they fire non vet probies, and maybe won’t even get that far if enough people take the DRP 2.0. I don’t know what will happen, but that’s my hopeful speculation. [/quote]

AI hallucination is a nonissue in this context. It hallucinates because it's allowed to do so. There are products that do not hallucinate, such as that provided by Lexis. The hallucination that you hear about in briefs and the news is because folks are using products like off the shelf GPT, which has no parameters that would prevent the hallucination. I was using AI before I joined BVA, and am a strong proponent for its rollout at BVA.

BVA is literally decades behind what the rest of the legal community is doing in terms of doc review, de-duplicating, predictive coding, analysis, etc. [/quote]

I’m hopeful that in the next 10 years, BVA will widely implement AI to adjudicate cases. If all goes well, BVA can reduce attorney headcount to 100 experienced attorneys to conduct quality reviews of AI decisions.[/quote]

Will AI make credibility determinations? That is a major issue that comes into factor with lay evidence.  Also, would CAVC accept a credibility determination made by AI?[/quote]

Sure, it could. It would consider the same things that we do - consistency, corroboration, testimonial infirmities, etc.

CCW can absolutely be programmed.

BVA work is very basic and formulaic. [/quote]

Actually, you don’t need judges then. You still need attorneys. That’s what your argument comes down to.  [/quote]

That is true. If you just rely on AI, you don’t need any judges. And I think you people are all making this way too simple. AI is only a tool. It can’t be an end. If I have something seriously wrong with me, I can turn to the Internet to help me diagnose, but I’m going to see a doctor for a real opinion. [/quote]

That’s why BVA needs a cadre of 100 or so experienced attorneys to review AI drafted decisions for quality. Humans won’t be completely out of the adjudication process, at least not yet.[/quote]

That’s not correct thinking. Unless you’re going to start from the position that AI is 100% correct, it’s going to take just as much time to review that decision written by AI and then compare it with the actual evidence of record. It might actually take more time.

It’s a lot easier to review the evidence and write a decision, then it is to review somebody else’s decision and then go through the evidence to find the evidence it supports. [/quote]

Agreed. People seem to think AI will just be correct. That’s not true at all. AI has been tested in courts; attorneys relied on AI to write briefs, and they have been disastrous. AI just makes up case names and fake citations. AI doesn’t have any control. It just gets input, and gives an output. [/quote]

AI is just a generic term for what generally refers to something that has a very, very large language database. There are countless AI tools. Some, like Chat GPT, don't have any parameters that would restrict hallucinating. Others, like Lexis AI, are confined to real law and cases and do not hallucinate.

AI can also be instructed to give citations and authority for any position it takes. It could cite to the record for any factual positions, just like a law journal does for virtually all propositions.

Chat GPT, for example can be given a set of your judge's cases and be told to rely on only those along with the CFR and USC and asked questions and told to cite to relevant authority for its positions. I know because I do just that. I update the dataset every few months. I can grab draft language out of it 10x faster than snippets. I can also give it fact patterns and it'll tell me how my judge has decided in the past and give me those cases. The searches are so much easier and faster than what we have on SharePoint.

If I were to put the effort into it, I could easily have it write decisions. Could it do the entire job? No, not as I have it implemented, but does it simplify my work? Oh very much so.

In terms of what the rest of the legal world is doing for doc review, it wouldn't be hard to have AI doing far more work.

Take 5 minutes and look at what platforms like Everlaw are doing.

BVA is in the stoneage with Case Flow. [/quote]

Options

	Disable HTML in this message
	Disable BB Code in this message
	Disable smilies in this message

Review message

Search Recent Topics Hottest Topics