Detecting Regulatory States in Natural Language: A Validation Study of the Four-Mode Gradient Framework

This study presents the first empirical validation of the Four-Mode Gradient, a theoretical framework proposing that human regulatory states can be classified into four modes—Connection, Protection, Control, and Domination—based on perceived safety or threat. Using natural language analysis of 10,000+ Reddit "Am I The Asshole" (AITA) posts, we tested whether these four modes could be detected using validated psychological constructs from polyvagal theory, attachment research, contempt markers (Gottman), and moral disengagement theory (Bandura). Key findings: - All four modes successfully detected in natural language - Mode classifications correlated with external community judgments - 33.8% of individuals escalated toward Control/Domination when challenged - 22.2% de-escalated toward Connection - Self-awareness (measured through "complexity markers") differentiated trajectories - De-escalators showed 78% higher complexity marker rates than escalators Critical discovery: The key variable is not current state but capacity to return to Connection when challenged. Pre-registration: https://osf.io/f4x6y Framework: https://teg-blue.com