Neurotransmitters could improve adaptability in artificial neural networks
Position Paper: Neurotransmitters as a Missing Dimension in Artificial Neural Networks
Neural and Evolutionary Computing
Summary
Artificial neural networks today learn by adjusting connections uniformly, but they miss how real brains use chemicals called neurotransmitters to change learning more flexibly. The authors argue that adding this chemical signalling as a new factor could help artificial systems learn better and keep knowledge longer. They suggest this idea offers a fresh way to design smarter and more adaptable AI, but they do not test or build such systems yet. This is a call to explore these biological principles in AI design.
What this means in practice
- •For ai system designers: Design neural networks that include neurotransmitter-like signals to improve adaptability and long-term stability beyond current methods.
- •For robotics engineers: Develop robots with neural control systems that mimic brain chemical modulation to enable more flexible and sustained learning behaviors.
A position paper. It proposes an approach and reports no results.
Authors
Yupei Li, Manuel Milling, Berrak Sisman, Björn Schuller
Abstract
Artificial neural networks (ANNs), as core components of modern deep learning (DL) systems, lack the adaptive flexibility and long-term stability exhibited by biological systems. This limitation largely stems from the fact that conventional ANNs rely on uniform, local, and gradient-based parameter updates, while neglecting internal learning principles that are biological mechanisms such as neurotransmitters signalling or neuroplasticity. Consequently, many existing approaches focus on architectural expansion or mathematical fine-tuning techniques such as regularisation or parameter isolation. Inspired by the superior adaptability and plasticity of mammalian brains, we posit that neuromodulation with neurotransmitters constitutes a third axis of learning, complementary to neural activity and synaptic plasticity, and should be explicitly modelled in artificial neural networks. In this positional paper, we argue that incorporating neuromodulatory principles into ANN design represents a promising and underexplored research direction, and we advocate for greater attention to this perspective in the development of adaptive and continual learning systems.