With the recent viral explosion of ChatGPT taking over the internet, it seems like discussions around AI are everywhere. What is it? How do you use it? How shouldn’t you use it?

And AI is top of mind for professionals—whether you are using it, creating AI-powered solutions, or considering it for your business. But the truth is, AI has been around for decades.

Do you have a Roomba? A smart thermostat? Netflix (how do you think it gives you those personalized recommendations…)? Or even the fitness mirror? The truth is, you may have been using AI longer than you realize.

There are different approaches to AI, and when it comes to hiring, the approach you use matters.

We sat down with Dr. Lindsey Zuloaga, Chief Data Scientist at HireVue, to discuss the different types of AI on the market, why static AI is the standard for hiring, and the red flags to look for when searching for a vendor.

Dr. Zuloaga explains that when people talk about AI today, usually they’re referring to a particular class of AI systems called Machine Learning algorithms. These systems are trained to predict an outcome based on patterns in previous data.

Dr. Zuloaga says that when we think about controlling AI systems, an important distinction is between dynamic and static AI. Dynamic AI involves learning new patterns according to the data coming in on the fly.

“With dynamic AI, there is feedback coming into the system, and the system is regularly updating. For example, think of a search algorithm that takes into account what people are clicking on and what seems to be higher quality. This data is then fed back into the algorithm to give you a result. You may get a different result from one minute to the next, because the algorithm is constantly learning and changing.”

Static AI works differently—as the name may suggest.

“With static AI systems, the algorithm is trained and tested in the ‘lab,’ and then locked before deployment. The system can only ‘learn’ new things if someone chooses to update it.”

Whether static or dynamic, AI systems can be complex and sometimes behave in unexpected ways. This is why the algorithms we use to assess candidates for jobs are not only static but are rigorously tested for validity (their ability to predict job-related competencies) and fairness (to ensure there are not significantly different outcomes among demographic groups). These tests occur before an algorithm ever interacts with a candidate, and they are updated no less than annually to ensure the highest standards of fairness.

What about Generative AI? Generative AI is a new frontier in AI that will have large scale effects.

According to Forbes, “Generative AI is capable of generating new data by recognizing patterns in existing data. In the context of conversational AI, it involves utilizing machine learning algorithms to produce natural language responses to user queries or requests.”

Dr. Zuloaga describes generative AI as a technology that makes something itself based on a prompt. Systems like ChatGPT are based on Large Language Models (LLMs) that are trained on vast amounts of text. Human feedback into the model can supply extra training data to optimize how the model responds in a conversational way.

Generative AI can be static or dynamic. ChatGPT is not being re-trained on the fly (it’s not dynamic), but that doesn’t mean you’ll get the same answer every time you ask it the same question (the variability is actually a setting you can tune in the system). Further, chat-based systems track an ongoing dialogue, so how they behave changes based on the information that was supplied throughout the conversation.

Standardizing static AI in hiring

At HireVue, we believe static AI that is deterministic (meaning it gives the same output every time given the same input), should be the standard of hiring.

Dr. Zuloaga emphasizes, “To adhere to laws and best practices in the hiring space, hiring assessments should be locked. Fairness is extremely important. No one should be using a different algorithm to evaluate different candidates moment to moment.”

She continues, “Human bias in hiring is a well-studied phenomenon. We are very careful about what data we take as feedback into our algorithms.”

HireVue is proud to be at the forefront of legislative action working to ensure all AI use is ethical. According to HireVue CEO Anthony Reynolds, there are 6 criteria that we want taken into account with new laws—and HireVue has been active in bringing them into every session as a framework.

Uniform audit criteria must be defined in the context of automated employment decisions tools
Hiring algorithms should be static and deterministic (rather than dynamically changing in the wild)
Audits must be required for all automated employment decision-making tools, not just those that use machine learning
Notice & transparency to empower candidates
Employers need to be obligated to supply demographic data
Vendors should deliver audits – this is too great a burden to place on employers

Hiring AI isn’t based on new science or theories—and works best with humans.

Static AI with repeatable results is the standard of HireVue’s technology, and while AI in hiring may seem to be a newer concept, the standard isn’t.

It’s important to remember that the methodology behind our technology may be newer, but the vast legal structures governing fairness in hiring have been around for a long time. We adhere to the law and follow best practices that go beyond those standards.

HireVue technology predicts competencies—not human hiring decisions.

Dr. Zuloaga reminds us, “What we’re trying to predict is a very well-defined job-related competency, not if someone like you was hired in the past. Our technology works in conjunction with humans, debunking the idea that it will make decisions by itself.”

HireVue believes that the best decisions are made when technology and humans can work together. “Humans should be using AI as decision support. We’re not taking over human discretion, but we do believe humans and tech working together is the best solution.”

AI should be explainable.

When it comes to building your HR tech stack, there are a lot of options out there, and weeding through the good, bad, and ugly can be overwhelming. Dr. Zuloaga encourages TA teams to think through the explainability factor when looking for a vendor.

“If a vendor cannot explain how their technology works in easy terms, you should be wary.”

Vendors should be able to explain how it works easily and explain it in a way that makes sense.

“They should also be able to share with you the validity and ROI of their product. If they can’t, it may be smart to look elsewhere.”

Want to learn more about the transparency behind HireVue solutions? Ready the industry’s first AI Explainability Statement here.

Customer Stories

Great Southern Bank