Reconceptualizing job crafting through machine learning with the construct mining pipeline

Abstract Job crafting encompasses diverse strategies and behaviors that vary across individuals and contexts, making its definition and classification difficult and subjective. Attempts to conceptualize job crafting have faced one of the greatest challenges associated with construct conceptualization: the difficulty in capturing the complex and multidimensional nature of the phenomenon. This study introduces a novel methodological approach, the Construct Mining Pipeline (CMP), combining natural language processing (NLP) and machine learning (ML) to refine the conceptualization of job crafting. By analyzing textual data from structured questions with a BERT-based model, we identified key dimensions using dimensionality reduction and clustering techniques. The analysis uncovered nine distinct dimensions of job crafting, revealing components not fully captured by traditional categories, such as technological optimization, task prioritization, and relational strategies. These findings highlight the multidimensional and dynamic nature of job crafting, broadening existing perspectives to include contemporary work realities such as digitization and collaborative dynamics. The CMP method demonstrates the potential of artificial intelligence (AI) to bridge qualitative and quantitative methodologies, providing a robust framework for advancing the understanding of complex psychological constructs.