Why All the AI Research You’re Reading is Dead Wrong

May 13, 2019

Why All the AI Research You’re Reading is Dead Wrong

When I was writing my book, Artificial Intelligence for HR, I thought very hard about doing a study on AI adoption by HR professionals. I wanted to understand the degree of penetration into today’s workplace, and the book was my perfect “excuse” as a researcher to explore the topic. However, in my digging I ran across half a dozen studies that were asking similar questions, and all of them had wildly different results. Still today I see new studies coming out that ask questions like:

  • Are you currently using AI to support your HR function?
  • Are you planning to buy machine learning solutions?
  • And so on…

The problem? Every week I’m presenting or speaking directly with HR leaders from around the world. For a sense of scope, I’ve already physically spoken in front of several thousand HR, recruiting, IT, and business leaders so far this year and have plans to speak to at least 5,000 more before the end of the year. And the truth is that more than 90% of them don’t have a firm grasp on what AI is and how it works.  This isn’t an indictment: we’re people professionals, not technology professionals, but it is illuminating in another way.

The Importance of Research and Survey Design

See, my team and I design research studies that are very carefully structured to understand market trends, buyer behavior, and other relevant topics. If those questions are not clear, the answer choices don’t make sense, or the audience doesn’t understand the topic, then the responses are garbage.

For example, we are validating a survey right now on compensation technology. In the testing phase we realized that one of the questions (which makes perfect sense to me) was not understood by those taking the survey, meaning the data we gathered for that particular question is totally useless. We decided to modify the question to allow a different, simpler type of response, but it just goes to show that survey design and understanding your audience is very, very critical if you want to have credible data.

Back to our topic today on the AI “research,” from a researcher’s perspective, the data I’m seeing means those people responding to surveys are making a lot of guesses but aren’t really sure what their responses even mean. In other words: garbage. Those gathering the data want to do a service and want to offer insights, but they are doing so at a cost of sharing inaccurate information.

External Validation

If my perspective isn’t enough for you as the leading voice on AI development and applications in the human capital space, I also spoke with a master’s degree student doing research on this topic for their capstone course. This student based in Europe was able to gather data points on this same topic and found similar levels of understanding in the UK and surrounding areas to what we have seen for the North American market.

We had a conversation this past week and compared notes, finding that (again) the average HR leader will respond positively about using AI but in subsequent questions fails to identify the solutions it’s implemented in or the method in which it supports their capabilities as an HR team.

Bottom line: this isn’t a local problem–it’s global.

The Forward Thinkers

I will admit that there are some HR and business leaders that have taken the steps to educate themselves (whether through the book and hearing me speak or of their own volition) on what AI is and how it works. But there’s still a gap: those that do understand the basics of AI still don’t know if their HR system uses it or not, meaning the data still isn’t worth your time to consume it (yet).

In the end, I don’t know how much it really matters other than the fact that i1) it’s incredibly annoying as a researcher 🙂 and 2) this is an encouragement for HR leaders not to be disheartened or feel behind when they see data from some “reputable” source saying that 80% of HR departments will use AI by 2022. That’s simply not true.

This space is so new that the average buyer doesn’t yet have a good handle on the technology and its capabilities yet. Be careful who you listen to and what you consume when it comes to this topic. And if you ever want insights from our research, don’t hesitate to ask.