Integration of Machine Learning Methods into the Sociological Analysis of Client Experience in the Legal Services Market

The article explores the possibilities and methodological foundations for integrating machine learning methods into the sociological analysis of client experience in the legal services market. Using the LegalBench learned_hands_consumer corpus (620 complaint texts), the authors conducted an empirical comparison of an interpretable model (TF‑IDF + logistic regression) and a neural network model (fine‑tuned Jina-embeddings with LoRA). Based on error analysis and the confidence distribution, four behavioral types of clients are identified. A four‑level analytical system is proposed, and practical recommendations are formulated for optimizing client experience, including a two‑stage ML filter, query segmentation, and a human‑in‑the‑loop retraining cycle. The results demonstrate that machine learning can serve not only as a technical classifier but also as a tool for sociological reconstruction of digital traces of legal mobilization.