Attorney: Even with AI, humans must run human resources

Listen to this article

Artificial intelligence has already made its presence known in the legal sphere. Lesser known is its presence in employment decisions such as recruitment and retention. Lisa Ellingson, attorney at Winthrop & Weinstine, explains how AI is already being used in these areas, unbeknownst to some users and some attorney employees.

First of all, Ellingson says, AI is already being used with respect to legal employment. “It’s not just for line workers or cashiers,” Ellingson explains. For instance, “Suited” is a talent screening platform that says it “fairly and accurately” assesses applicant’s professional potential.

“Nobody is going to be above or below the use of AI in employment,” Ellingson says.

Ellingson states that AI has been used in a variety of ways with regard to employment decisions. One way is resume screening, where AI can do a “first pass” or a “final pass” through a pile of resumes. It might scan a resume for certain keywords, prioritizing resumes that have certain keywords in them, or the right frequence of keywords. AI is also able to look at statistical data in resumes such as timeline gaps that might be concerning.

Many interviews are now conducted by AI. AI will analyze people’s physical responses, gestures, and even their vocal responses. “That’s actually surprisingly common,” says Ellingson.

“We don’t just see this playing out at only the biggest companies. Even smaller companies are using some of this tech,” Ellingson says. “They’re really gaining in popularity.”

A lot of companies are using these tools for efficiency purposes. “These days, HR departments will get hundreds of resumes for a position, maybe thousands of resumes. It’s just not possible for some of these HR departments to really meaningfully go through those,” Ellingson says.

Additionally, AI is also being used to monitor productivity, providing information about customer feedback and online behavior, and bundling that into a performance review.

“I don’t think it always is communicated,” Ellingson says of AI tools that monitor employee behavior. “A lot of states do not have strong employee privacy requirements.”

Employers cite benefits of AI tools

Ellingson explains that some people are finding that they can get some standardization out of these AI tools. They are providing additional insights to be able to analyze these metrics that people otherwise would not be able to compile and cut through as well. She thinks that this can be a good thing.

“I think there is potential benefit to having more standardized systems, a standardized program that you can cite to that is auditable, because, again, the existing system is people, who can make all sorts of subjective choices that are not auditable,” Ellingson states.

Having AI to point to can also eliminate some risk when it comes to firing, especially. Citing why a person should be terminated can be nebulous, something that having metrics to point to can eliminate. “It’s not just because there was some office politics or that sort of thing that is hard to quantify,” Ellingson explains.

Ellingson also suggests that AI can help with increasing diversity, something that many organizations are continuing to work on.

“The alternative to AI is just the human system that is already in place,” Ellingson asserts. “If you have an HR department full of people who are struggling to improve the diversity situation, maybe using an AI tool will help them to improve their metrics.”

AI can invite legal problems

One problem with AI is that some people are using AI without even knowing that they are doing so. Many companies also may be using tools where AI is rolled into them, but the employers are unaware that AI is present.

“One of things that we recommend is evaluating any new HR tool that you are going to use, because you don’t always know that there is AI,” Ellingson says. “And, if there is AI, you have these extra issues and, depending on the state that you’re in, you have additional legal obligations and client obligations. You need to know this from the outset.”

Another problem with AI, in the employment context as well as others, is the “black box” problem, where we are unaware of how AI arrives at the outputs it produces. “Once you decide that you are going to use it, you have to have an understanding of how it works,” Ellingson notes.  Ellingson says that this is hard because of the black box problem.

In states such as Colorado and Illinois, employers will be required to describe, if asked, how the AI arrives at its decisions. “You can’t do that if you don’t know, or have some idea,” Ellingson explains.

In Minnesota, there is the new Minnesota Data Privacy Act. Ellingson says that it is a little bit unclear about how it will apply to employment, as there is some ambiguity within the law as written. She asserts that one section seems to apply to employment, while another seems to suggest it does not apply to employment.

“We’re really looking at whether that may impact employment as well, because it does give a data subject the right to review any profiling that has been done about them and be informed about what actions might have been taken for a different decision. So, it is possible that AI in employment decisions could be covered under that as well,” Ellingson says. Ellingson suggests that, as there has already been an explosion of AI-specific legislation, more is on the horizon.

On the front end, Ellingson says that applicants and employees need to know that employers will be using AI on them and get their consent. “It really reduces legal risk for compliance issues, and risk that people will come back really unhappy about that,” Ellingson explains.

Humans are still needed, at this point

While AI is making its presence known, in employment broadly but also in the legal sphere specifically, Ellingson stresses that humans are not yet redundant in these processes. Ellingson says that employers can let the AI tool help them with the function, but that a human must be involved to actually make a decision.

“There’s a lot of risk if you are just going to let the machine decide who is going to be hired or be fired. Maybe the machine can help inform you about certain metrics or whatnot, help cut the data in a different way, but, ultimately, get a human involved and have them review the results, have them make the decision,” Ellingson cautions.

“You really need to have humans involved,” Ellingson avows.

RELATED:

AI shakes up med-mal law

ABA panel offers guidance on generative AI tools

AI’s big advances yield challenges for clients navigating benefits and risks