Key points
- AI uncovers suboptimal, human-like behavioural patterns
- Traditional decision models miss real-world behaviour
- Individual strategies show potential for mental health insights
ISLAMABAD: For years, researchers have been intrigued by how humans and animals make decisions, particularly through trial-and-error learning based on recent experiences.
But traditional models of decision-making may not fully reflect reality. These frameworks often assume that we make optimal choices by carefully weighing our past experiences—a notion that might oversimplify how decisions are actually made.
Now, a newly published study offers a fresh perspective by using artificial intelligence in a novel way to better understand this process, according to Neuroscience news.
Rather than relying on large, complex AI systems, the researchers used tiny artificial neural networks—small enough to be clearly understood, yet still capable of capturing complex patterns of behaviour. Their findings reveal what truly drives individual choices, even when those decisions are not the most efficient or logical.
Rethinking how brains decide
“Instead of assuming how brains should learn in order to make the best decisions, we developed an alternative approach to discover how individual brains actually learn,” explains Marcelo Mattar, an assistant professor in New York University’s Department of Psychology and one of the paper’s authors. The study appears in the journal Nature.
Model overview and performance on animal tasks
“Our method acts a bit like a detective, uncovering the real decision-making strategies used by animals and people. By working with very small neural networks, we have uncovered patterns that conventional models have missed for decades.”
The research highlights that these streamlined neural networks—simplified versions of those used in mainstream AI—are significantly better at predicting animals’ choices than traditional cognitive models. That is largely because they account for the fact that behaviour is often suboptimal, rather than perfectly rational.
Deep understanding
In controlled lab settings, these small networks proved just as accurate as much larger, commercial-grade AI systems in predicting choices. But because they are simpler, they allow researchers to examine and understand the mechanisms behind decisions in a much clearer way.
Model performance using knowledge distillation
“Using these tiny models makes it easier for us to apply mathematical tools to interpret what’s going on in someone’s mind when they make a choice,” says Ji-An Li, a PhD student in the Neurosciences Graduate Programme at the University of California, San Diego.
“Big neural networks, like those used in commercial AI, are brilliant at making predictions—for instance, guessing which film you might want to watch next,” adds co-author Marcus Benna, an assistant professor of neurobiology at UC San Diego’s School of Biological Sciences. “But it’s much harder to explain why they’ve made that particular prediction.”
“By training the simplest possible versions of these AI models to mimic animals’ choices—and then analysing them using tools from physics—we can understand how they work in a more transparent and meaningful way.”
Rethinking rational behaviour
Understanding how learning from experience shapes decision-making is not only a key goal in neuroscience and psychology, but also has wider applications in fields like business, technology, and public policy.
However, the new study argues that conventional models—focused on ideal, optimised decisions—often miss the messy, imperfect nature of real-world behaviour.
Dynamical systems analyses for interpretation and comparison of multi-dimensional models
Strikingly, the new model successfully mirrored the decision-making styles of humans, monkeys, and lab rats. It even predicted the kinds of less-than-perfect decisions people and animals often make, making it a better reflection of actual behaviour than previous approaches.
What is more, the model could predict individual-level differences—revealing how each participant used unique strategies to make their choices.
“Just as studying individual physical differences has reshaped modern medicine, recognising how people differ in their decision-making styles could lead to breakthroughs in how we understand mental health and cognitive function,” Mattar concludes.
