Quantum reinforcement learning in human decision making
Published by Nature Human Behaviour, it is the first time quantum reinforcement learning has been empirically tested on human decision making.
UNSW Canberra Associate Professor Daoyi Dong instigated the concept of quantum reinforcement learning (QRL), which takes advantage of the unique characteristics of quantum computation.
It builds on existing research of classical reinforcement learning (CRL).
“CRL is a machine learning method that has also been widely used in decision psychology,” Associate Professor Dong said.
“It addresses how an agent (a computer or an animal) will learn to make decisions based on the eventual reward or penalty.”
Quantum computation, which processes information faster than the classical binary code, has already been successfully applied in the field of machine learning, including robot navigation.
The researchers’ goal was to explore the quantum-like aspects of processes underlying value-based decision making.
Behavioural and functional magnetic resonance imaging (fMRI) data was collected from 58 healthy subjects and 43 nicotine addicts performing the Iowa Gambling Task (IGT),”Associate Professor Dong said.
The IGT is a value-based decision-making task that is well-known among psychiatrists and neuroscientists, designed to evaluate the degree of decision-making defects.
Smokers were selected as they are reported to show impaired decision making in the IGT. This further validates the conclusions about the mechanisms of decision making in healthy people.
“The data was examined using two quantum reinforcement learning models and 12 classical reinforcement learning models,” Associate Professor Dong said.
The paper concluded that in all groups “the QRL models performed well when compared with the best CRL models and further revealed the representations of quantum-like internal-state-related variables in the medial frontal gyrus in both healthy subjects and smokers”.
This suggests that value-based decision making can be illustrated by QRL at both the behavioural and neural level.
“These findings are useful to understand value-based decision-making at both the behavioural and neural levels, and also provide the fMRI evidence connecting quantum cognition to neuroscience,” Associate Professor Dong said.
He said the next phase of research in this area is to provide theoretical and experimental evidence for quantum-like process during human decision making.
Quantum reinforcement learning during human decision-making was published in January by Nature Human Behaviour.
The research team was led by Xiaochu Zhang from the University of Science and Technology of China, and also included researchers from the Shanghai International Studies University, Shanghai Jiao Tong University, the RIKEN, the University of Michigan, the Anhui Mental Health Centre and the Tianjin Normal University