Is ChatGPT making the world fall victim to the Dunning-Kruger effect?

It’s all the rage. Really. Endless posts and articles about how AI, and ChatGPT (GTP-3) in particular, will revolutionize the world and make us all more productive.

As a trusted advisor and consultant in IT, I find myself in the nexus of complex technology and people every day. And let me tell you: ChatGPT is already becoming a real pain that can significantly reduce team performance.

Why you ask? Because ChatGPT seems to amplify the Dunning-Kruger effect. If you haven’t heard about it, the Dunning-Kruger effect is a cognitive bias that can lead you to overestimate your abilities and knowledge in a particular area, even if you have limited or no experience. It's well documented and very real.

When you interact with ChatGPT, you might be tempted to trust its responses without questioning their accuracy or validity. This can be especially true if you are discussing a topic that you are unfamiliar with. For instance, if you are talking about a complex scientific theory, you may trust ChatGPT's responses without questioning them simply because you do not have the knowledge or experience to evaluate them. You also may assume that you can handle new areas of expertise simply by running through a few ChatGPT sessions. It can really make you look like an expert in a meeting when everyone else knows even less about the topic.

Run this experiment: Assume you’re not working with IT security to begin with, but you’re about to attend a meeting about the controls in the CIS v8 framework. Start a new ChatGPT session, and enter something like this into the prompt: “I am about to attend a meeting on IT security where CIS v8 will be discussed. What low-hanging fruits specifically from the CIS v8 framework could we focus on?”. Voila, you now sound like an expert. If you have more than five minutes to prepare, you can even ask it to elaborate on each of the suggestion it spews out. With half an hour you could probably arrive with a 7-page draft Word document and a PowerPoint presentation.

So, now you sound like someone who knows what they’re talking about, in the context of CIS v8, but you really know little more than what an advanced chatbot gave you in a minute-long session. This presents two problems. First, it makes you assume you can do a lot of things totally outside your league, because it presents you as competent in an area you know nothing about. This can lead you or the team to some bad decisions. Second, it makes it hard for others around you to accurately asses your skills.

Now try a new experiment: Imagine you are an IT Operations manager in a small-medium sized business. You have a meeting with your Operations staff about using the CIS v8 framework. One of your employees seems to know a lot about the topic already, and offers low-hanging fruits you can start with. He even brings a short presentation, with the most important highlights. Do you:

a.) Praise them for being knowledgeable, taking initiative, and involve them in the process?

b.) Call it out, and ask them if they used ChatGPT to make themselves sound smarter than they are?

c.) Leave it be, because you assume/hope they will perform great even if they just used ChatGPT for a few minutes before entering the meeting?

I am not going to make this an actual poll, and I can’t you tell how widespread this is. But I can tell you it’s happening within my field of expertise already. I can also tell you that detecting it and calling it out will not necessarily win you any points. So if you even think that this is happening to you, you are probably check mate already.

To avoid falling prey to the Dunning-Kruger effect when you are using ChatGPT yourself, you need to maintain a healthy dose of skepticism and critical thinking. You should always question the accuracy of ChatGPT's responses and seek out additional sources of information to confirm or refute them. That’s ChatGPT’s own response by the way. Let that sink in for a moment... ;-)

But how can we avoid falling prey to other people’s ChatGPT-imposed Dunning-Kruger effect? That is a larger question that’s going to be harder to answer. And off course they don't even know they have fallen victim to a cognitive bias - that's the very definition of it. So unless you're an expert diplomat, you will probably hurt their feelings or even anger them if you call it out.

I don’t have a clear answer to this problem. But I welcome your advice in the comments, because I need help tackling this. Perhaps we all do.