Tensor Networks Model Children's Emotional Memory with 77.98% Accuracy

A team of researchers (Groves, Jackson, Robertson, Hance) has published a paper on arXiv demonstrating how tensor networks—mathematical structures often used in quantum physics—can model the order-dependent nature of emotional memory in children. In their study, children were shown sequences of toys with different emotional valences (positive, neutral, negative) and then asked to recall the order. Standard psychological models, which treat each toy's recall independently, performed poorly because they couldn't capture how the valence of one toy influences memory of the ones before and after. The researchers introduced a classical tensor network that factors in this contextual dependence.

This tensor network model achieved a 77.98% accuracy in predicting children's recall sequences, a massive improvement over existing models. While the approach is not strictly "quantum cognition"—it runs on classical hardware—it draws inspiration from quantum mechanics' ability to represent complex dependencies. The paper argues that quantum-inspired methods like tensor networks are uniquely suited for modeling order-dependent phenomena in psychology. Beyond the accuracy gain, the study introduces a new experimental protocol for exploring emotional temporal memory, which could be analyzed using both classical and quantum-like cognitive models. This work bridges machine learning, neuroscience, and quantum physics, opening doors for more realistic cognitive modeling in real-world settings.