“Are We Still in Control?”: Exploring Patterns of AI Dependency in Scientific Research

The growing use of artificial intelligence (AI) in scientific research has raised concerns about “AI dependency”, a phenomenon that remains conceptually ambiguous and underexplored. Guided by self‐regulation theory, this study proposes a four‐quadrant typology of AI dependency based on goal orientation and self‐efficacy. Semi‐structured interviews with 20 researchers revealed four distinct patterns: collaborative active, instrumental active, passive compensatory, and passive pathway. Researchers with high goal value and high self‐efficacy (collaborative active) treat AI as a knowledge collaborator while maintaining autonomy. Those with high self‐efficacy but low goal value (instrumental active) prioritize efficiency and treated AI as a pragmatic tool. In contrast, those with high goals but low self‐efficacy (passive compensatory) relied on AI to compensate for skill gaps, while individuals low in both dimensions (passive pathway) exhibited habitual dependence and emotional distress when AI was unavailable. These findings reveal the complex psychological and behavioral dynamics underlying AI dependency, offering a more nuanced conceptual understanding and informing interventions that promote critical, self‐regulated AI use.