WCP 2025

Introduction This research aims to integrate artificial intelligence technology, particularly automated machine learning methods, into psychological research to address the limitations of traditional frequentist statistical approaches in handling nonlinear relationships of psychological characteristics. Current psychological research predominantly relies on regression-based methods, which often fall short in capturing the complex nonlinear relationships between psychological features. We propose an innovative semi-automated workflow that empowers psychology researchers to leverage machine learning algorithms for intelligent model selection, facilitating the construction of more precise and insightful theoretical frameworks.

Methodology We designed a low-code workflow based on AutoGluon, specifically tailored for psychological research methodologies with high accessibility. This approach encompasses three primary research objectives:
1. Automated hyperparameter tuning to attain optimal models.
2. Identification of important features through interpretability techniques, facilitating feature selection based on calculated importance.
3. Theoretical reconstruction based on important features by integrating exploratory factor analysis with machine learning interpretability.
Using psychological resilience research as an example, we provide a detailed annotated code workflow along with raw data to demonstrate the application of this method.

Results The workflow successfully implements automated machine learning model selection and optimization, effectively identifying key features in psychological resilience research. Through interpretability techniques, we quantify the relative importance of different features, providing a data-driven basis for feature selection. Combined with exploratory factor analysis, we reconstruct theoretical frameworks that more accurately reflect the nonlinear relationships among psychological characteristics.

Conclusion The semi-automated workflow proposed in this study provides psychological researchers with a practical tool to better understand nonlinear relationships between psychological characteristics. This low-code approach lowers the barrier for psychological researchers to utilize advanced machine learning techniques, offering new possibilities for data-driven psychological theory construction. Future research can extend this methodology to additional psychological research domains and integrate more types of data, such as text, audio, and images.