Tiny ai models reveal how we really make decisions

Summary: decision making often implies proof and error, but conventional models assume that we always act optimally depending on the past experience. A new study used small and interpretable artificial neuronal networks to discover how humans and animals really make decisions, revealing the suboptimal strategies that we often use.

These models predicted individual decisions with greater precision than traditional theories by reflecting the imperfect behavior of the real world. This work could transform the way we understand cognitive strategies and adapt mental health or behavioral interventions.

Key facts:

Realistic insights: small AI models revealed that decision -making strategies are often suboptimated but systematic. Individual differences: the models predicted individual behavior better than the frames based on optimization. Most important impact: the results can inform mental health approaches by mapping cognitive diversity.

Source: NYU

Researchers have been interested for a long time in how humans and animals make decisions by focusing on the behavior of proof and error informed by recent information.

However, conventional frameworks to understand these behaviors can ignore certain realities of decision making because they assume that we make the best decisions after taking into account our past experiences.

A study recently published by a team of scientists displays AI in an innovative way to better understand this process.

By using small artificial neuronal networks, the work of researchers illuminates in detail which drives the real elections of an individual, regardless of whether these options are optimal or not.

“Instead of assuming how brains should learn by optimizing our decisions, we develop an alternative approach to discover how individual brains learn to make decisions,” explains Marcelo Mattar, assistant professor in the Department of Psychology at the University of New York and one of the authors of the article, which appears in the journal Nature.

“This approach works as a detective, discovering how animals and humans make decisions. Through the use of small neural networks, enough to be powerful enough to capture complex behavior, we have discovered decision -making strategies that scientists have overlooked for decades.”

The study authors point out that small neuronal networks, simplified versions of the neural networks typically used in commercial applications of AI, can predict the elections of animals much better than classical cognitive models, which assume optimal behavior, due to their ability to illuminate subophant behavioral patterns.

In laboratory tasks, these predictions are also as good as those made by larger neuronal networks, such as those promoted by commercial applications.

“An advantage of using very small networks is that they allow us to implement mathematical tools to easily interpret the reasons or mechanisms, behind an individual’s choices, which would be more difficult if we had used large neuronal networks such as those used in most AI applications,” adds author Ji-An, a doctoral student in the postgraduate program of Neurosciences at the University of California, San Diego.

“The great neural networks used in AI are very good to predict things,” says author Marcus Benna, assistant professor of Neurobiology at the School of Biological Sciences of UC San Diego.

“For example, they can predict which film would like to see below. However, it is very difficult to succinctly describe what strategies these complex automatic learning models to make their predictions, such as why they think that he will like one more than another movie.

“When training the simplest versions of these AI models to predict animals of animals and analyze their dynamics using physics methods, we can shed light on their internal work in more easily understandable terms.”

Understanding how animals and humans learn from experience to make decisions is not only a main objective in science, but, more widely, useful in business, government and technology fields.

However, the existing models of this process, because they are aimed at representing optimal decision making, often fail to capture realistic behavior.

In general, the model described in the new study of nature coincided with the decision -making processes of humans, non -human primates and laboratory rats.

In particular, the model predicted decisions that were suboptimal, which better reflects the nature of the “real world” of decision making, and in contrast to the assumptions of traditional models, which focus on explaining optimal decision making.

In addition, the Nyu and UC San Diego scientists model could predict decision -making at the individual level, revealing how each participant displays different strategies to reach their decisions.

“Just as studying individual differences in physical characteristics has revolutionized medicine, understanding individual differences in decision -making strategies could transform our approach to mental health and cognitive function,” concludes Mattar.

FINANCING: The investigation was supported by subsidies of the National Science Foundation (CNS-1730158, ACI-1540112, ACI-1541349, OAC-1826967, OAC-2112167, CNS-2100237, CNS-21120019), Kavli Institute for Brain and Mind, the University of California From the office of the office of the office of the Office of the UC San California San Diego San is Caliatuatana San Antigues San Calatra. Telecommunications and information technology/Qualcomm Institute.

About this AI and decision -making news news

Author: James Devitt
Source: NYU
Contact: James Devitt – Nyu
Image: The image is accredited to Neuroscience News

Original research: open access.
“Discover cognitive strategies with small recurrent neuronal networks” by Marcelo Mattar et al. Nature

Abstract

Discover cognitive strategies with small recurring neuronal networks

Understanding how animals and humans learn from experience to make adaptive decisions is a fundamental objective of neuroscience and psychology.

Normative modeling frameworks, such as Bayesian inference and reinforcement learning, provide valuable information about the principles that govern adaptive behavior.

However, the simplicity of these frames often limits their ability to capture realistic biological behavior, which leads to artisanal settings that are prone to the subjectivity of the researcher.

Here we present a novel modeling approach that takes advantage of the recurrent neural networks to discover the cognitive algorithms that govern biological decision making.

We show that neuronal networks with only one or four units often exceed classical cognitive models and coincide with the largest neuronal networks to predict individual animals and human options, in six learning tasks of well -studied rewards.

Critically, we can interpret the trained networks using dynamic systems concepts, allowing a unified comparison of cognitive models and revealing detailed mechanisms underlying the behavior of choice.

Our approach also estimates the dimensionality of behavior and offers information on the algorithms learned by artificial intelligence agents of meta-reference learning.

In general, we present a systematic approach to discover interpretable cognitive strategies in decision -making, offering information about neuronal mechanisms and a basis for studying healthy and dysfunctional cognition.