“We are all going to lose our jobs and AI will make humans extinct.”
Such are frequent, admittedly carricatured, prognostications by computer science PhDs, professors, Tech CEOs and even Nobel prize winners. These are perceived, rightfully, as some of the most intelligent people in the world. But as much as they know about technology, science and mathematics they show no deep understanding of things such as human behavior, psychology, sociology, labor market economy, technological history, biological processes that would play crucial roles in their doomsday visions.
There is a tendency for technologists to consistently overestimate their understanding of society and psychology because engineering success creates false confidence in other domains such as human behavior. Psychological research points to a few reasons why that might be.
The transferability myth
Transferability bias is when you think that competence in one area transfers to other unrelated domains. in Self-Insight: Roadblocks and Detours on the Path to Knowing Thyself. Psychology Press David Dunning argues that smart people are often more susceptible to this bias because they are able to elaborate justifications for their incorrect beliefs, a process known as motivated reasoning. In the article Cognitive sophistication does not attenuate the bias blind spot, the authors found that people with higher cognitive ability exhibit a bias blind spot that makes them disregard their own biases.
The beginner’s bubble
When you start out in a new domain there is evidence of an increased sense of competency in the beginning. Dunning continued digging into this and in Overconfidence among beginners: Is a little learning a dangerous thing? He and co-author Sanchez found that confidence does not rise linearly with knowledge. Instead, there is a bubble of exuberant overconfidence that happens after just a few learning experiences. So, as long as you don’t dig too deep into an area you will continue to feel overconfident.
The Engineer’s fallacy
The engineers fallacy is assuming everything can be thought of as a machine. Booker and Gates argue in their 2021 paper Cognitive Biases and the Cultural Disconnect between Engineers and Decision-makers that technical experts often apply closed system logic (engineering) to open-system problems such as those in sociology or economics, which leads them to a specific type of overconfident misjudgment.
We should therefore be wary of accepting any non-technical statements about AI just because people have deep technical knowledge, because there is an increased risk that first, they don’t know that they don’t know enough about the things they talk about and secondly, that there is an aura effect where perception of their high competence lends increased credence to their claims. After all, how much does the computer science elite know about normal peoples jobs, the job market, human intelligence and psychology? In my personal experience, the same as Johnny nobody next door.
Anyone can participate in public debates about society, technology and the future but there is no reason to pay special attention to the computer scientists just because they know the code, in fact there is increased reason to be suspicious of what they are saying.
Photo by Fer Troulik on Unsplash