Artificial intelligence (AI) extends human capabilities, helping to identify patterns in medical images, analyze large volumes of data, program, translate, and accelerate intellectual tasks. This capability supports a reasonable thesis: a society that learns to use AI can produce more knowledge and extend skills previously reserved for specialists. The challenge arises when efficiency begins to be confused with authority.
The psychology of human-machine interaction has been aware of this problem since before the rise of generative AI. It's called "automation bias" and consists of the tendency to place excessive trust in AI's results, even when faced with seemingly contradictory information. Therefore, a paradox arises: the more useful a tool appears, the greater our willingness to accept its errors as well.
An experiment published in Radiology In 2023, a study conducted by Thomas Dratsch and colleagues asked 27 radiologists to evaluate 50 mammograms with the assistance of a purported AI system. When the recommendation was correct, participants were right about eight out of ten times; however, when the suggestion was deliberately incorrect, the correct answers dropped to 19,8% among the least experienced, 24,8% among those with intermediate experience, and 45,5% among the most experienced specialists. Even expert knowledge proved vulnerable to a machine presented as competent.
Thus, value arises when the machine complements human judgment, while risk increases when it uncritically replaces it.
The crucial skill of our time is learning to question methodically, verify sources, reconstruct calculations, identify assumptions, and distinguish "appearance evidence" in AI-generated data. Hinton, the Nobel laureate who has warned about the inherent development of AI, has emphasized the need for caution, something that also applies to everyday risks.
The key lies in vigilant cooperation, striving for increasingly capable machines and, simultaneously, for competent citizens to examine them. Mature technology requires better-trained users, responsible institutions, and verification mechanisms.
Progress and prudence can go hand in hand.
Scientific caution is not about fearing the future, but about showing respect for the consequences. A truly technological society will be one that knows how to harness the intelligence of its machines and strengthen the judgment of its citizens.
@betancourt_phd
