More on AI in Labor Arbitration
I have written about AI before. In that July 2025 post, I opined that AI wouldn't make deep in-roads in the field of labor relations because it involves human relations even if the legal field could find some uses for it. I co-authored an article with Arbitrator Robert Creo about the ethical considerations of using AI. But I've continued to be reluctant to adopt AI for any purpose, even administrative ones.
Varsha Bansal
This has largely been for ethical reasons, although I acknowledge that LLM indexing will likely have positive scientific benefits. I know several arbitrators who have explored, and enjoyed, using AI tools for meeting management and note taking.
But more and more, I am avoiding AI for intellectual and stylistic reasons as well. Unfortunately, I see arbitrators who are using AI to research foundational labor law and labor relations questions, and the answers that the AI tools provide are inferior the work product of those same arbitrators. I do not understand why people who are experts in their field would turn to a glorified search engine to provide answers they are expected to know and be able to articulate easily.
In fact, AI is worse than a search engine for this sort of question, because rather than return primary and secondary sources the way a search engine would, it guesses at the answer after scraping all possible answers. It therefore does not distinguish among a pro-labor press release, an anti-union opinion, and legal decision. And it isn't even aware of its own bias. I know that the expert arbitrator can read it and sort out which parts might not be accurate on those topics they have knowledge of, but do they rigorously question the AI answers where they aren't familiar with the source material?
And what about the lay user - the grievant who is challenging their termination? Will they know whether the answer ChatGPT gives them is accurate? Unlikely. If ChatGPT give a different answer than the grievant's union rep, can the grievant bring a duty of fair representation claim against their union? ChatGPT will almost certainly tell them they have a winning claim.
Fortunately, a court recently held that an AI query is not an attorney-client privileged communication, strongly suggesting that the courts will not consider an AI determination of a grievance to be legally sound.
A friend of mine who does policy consulting told me that a client recently informed her that ChatGPT contradicted her advice, so they would be following the ChapGPT suggestion and terminating her contract. She told them that the AI suggestion actually ran afoul of law (it suggested befriending a state employee to get what they wanted rather than following the publicized process) and that they should consult their attorney before proceeding. I find this horrifying, and would find it more horrifying if my friend were the one using AI to generate ideas for her client.
Which is why I am so troubled that arbitrators and lawyers would use AI to research areas that are in their expected area of expertise. Perhaps unfairly, I find myself doubting those who use it for this purpose. Aren't you the expert? If so, tell us what you think, not what ChapGPT scraped from the internet.
