Labor Relations and AI

Today, I got an email offering me $60 an hour to provide my expertise on labor relations to a company designing an AI Human Resources program. Based on my per diem, my hourly rate is closer to $350 an hour, but even at that amount, I still wouldn’t provide content for a large language model (LLM) for AI HR. The human part is right there in the name.

I was going to make a joke about how an HR chatbot wouldn’t cheat with the CEO of their company, but then I remembered that people are now marrying their chatbot companions, so I decided not to make the joke.

Many lawyers and arbitrators are trying to figure out how AI will be a factor in their work. There have been lots of high-profile examples of its misuse, including lawyers who rely on fake citations and even a recent judge who had to pull back a decision that may have been AI-generated. But many more arbitrators and lawyers are testing the AI tools that are being introduced through commonly used software like Zoom and Microsoft. My colleague Robert Creo and I have an article about AI use coming out soon. I’ll share the citation when it’s published.

I don’t think AI will make any serious inroads into labor arbitration. Anecdotally, labor and management advocates vehemently oppose its use. An arbitrator’s human (or, in some cases, superhuman) understanding of human nature and the values, politics, and culture of a workplace is the reason that the parties use third-party neutrals. One union lawyer said he’s considering recommending that his clients negotiate language prohibiting the use of AI in dispute resolution. AI might be good for summarizing meetings and creating action items, but it doesn’t have the judgment needed to resolve workplace disputes.

Here’s an important thing to consider about AI: it is not more than the sum of its parts. It cannot have its own experience outside of its own LLM, and it can’t contribute new knowledge to itself. It relies on humans to provide additional information. Workplace dynamics are too subtle, too shifting, too subjective. They simply cannot be captured and fed into an LLM to a sophisticated enough degree to supplant experienced labor arbitrators. Someone might ask AI to generate a decision based on a database of arbitration awards, but AI doesn’t know if the witnesses are credible or whether the parties are barely containing their hatred for one another. The variety of human factors makes labor relations too HUMAN for AI to ever replicate, and I don’t think many people in labor relations want it to.

Stayed tuned for my other writing on this topic and, in the meantime, let me know what you think about AI use.