Decoding Artificial Intelligence: A Tutorial on Neural Networks in Behavioral Research

• Published in: Clínica y salud Vol. 36; no. 2; pp. 77 - 95
• Main Authors: Martínez-García, Javier, Montaño, Juan José, Jiménez, Rafael, Gervilla, Elena, Cajal, Berta, Núñez, Antonio, Leguizamo, Federico, Sesé, Albert
• Format: Journal Article
• Language: English
• Published: Colegio Oficial de Psicólogos de Madrid 01.07.2025
• Subjects: artificial intelligence; artificial neural networks; backpropagation algorithm; behavioral and health sciences; multilayer perceptron; python
• ISSN: 1130-5274; 2174-0550; 2174-0550
• DOI: 10.5093/clh2025a13

- Simplifying Complex Concepts: This tutorial helps to demystify ANNs by breaking down the backpropagation algorithm into manageable steps. Readers will gain hands-on experience in Python, empowering them to confidently replicate analyses for regression and classification tasks without feeling overwhelmed. - Building Confidence in Application: Designed for behavioral scientists, and even for other disciplines, this tutorial bridges theory and practice, alleviating anxiety around complex models. Learn to interpret results clearly and effectively, fostering a supportive environment for innovative applications of ANNs in research and beyond. Background: Artificial Neural Networks (ANNs), particularly multilayer perceptrons (MLPs) with backpropagation, are increasingly used in Behavioral and Health Sciences for data analysis. This paper provides a comprehensive tutorial on implementing backpropagation in MLP models for regression and classification tasks using Python. Method: The tutorial guides readers step-by-step through building a backpropagation MLP using a simulated data matrix (N = 1,000) with psychological variables, demonstrating ANNs’ versatility in predicting continuous variables and classifying (binary and polytomous) patterns. Python scripts and detailed output interpretations are included. Results: MLP models trained with backpropagation show effectiveness in regression (R² = .71) and classification (binary AUC = .93, polytomous AUC range: .81-.93) on test sets. Conclusions: This tutorial aims to demystify ANNs and promote their use in Behavioral and Health Sciences and other fields, bridging the gap between theory and practical implementation.