AI and the Dunning–Kruger Effect
This column was originally written in Dutch. This is an English translation.
By Bob Homan, Head of ING Investment Office
I recently read about the Dunning–Kruger effect, and I recognized it immediately. This effect refers to the tendency of people with limited knowledge to underestimate the complexity of experts' work. I sometimes catch myself doing this, and perhaps you recognize it as well.
When you connect this effect to the use of AI, it is easy to imagine what can happen. Reality may be oversimplified, leading companies to replace tasks, employees, or even entire departments with AI. To be clear, I am certainly not opposed to the use of AI. However, if you are in a position to make such decisions, it is worth being aware of the Dunning–Kruger effect. If AI takes over a task that turns out to be more complex than expected, the quality of both the task itself and the resulting product or service is likely to decline. If this happens on a larger scale—which is by no means unlikely—we could see the opposite of what we have become accustomed to throughout our lives: quality deteriorating instead of improving.
A brief economics lesson illustrates the point. This would represent a break from the long-standing assumption that inflation tends to be overstated because official measures often fail to fully account for improvements in quality. Consider a laptop, for example. Every year it becomes faster and more powerful, yet its price does not increase by very much. If inflation is overstated, then real economic growth is automatically understated. However, if AI-produced goods and services are of lower quality, the opposite occurs. Inflation is understated because consumers receive less value for the same price. As a result, real economic growth is overstated.
A similar principle applies at the company level. If businesses deploy AI indiscriminately and the quality of their products or services declines, their profits are also likely to suffer. Customers will notice the deterioration in quality, demand lower prices, or switch to competitors. As an investor, you can take both the macroeconomic and the company-specific effects into account. For now, the latter seems the more practical consideration. Investors and analysts can assess how effectively companies are implementing AI. Companies whose managers fall victim to the Dunning–Kruger effect may be best avoided.