Andrew Roetman
Fellow since Feb 2026
Andrew Roetman is a PhD Candidate in Clinical Psychology in the Gordon F. Derner School of Psychology, Adelphi University. His dissertation project is titled “Modernizing Implicit Motive Assessment: Training and Validating LLMs for Automated TAT Scoring and Prediction.” Andrew serves as a graduate research assistant in the Laboratory for Unconscious Processes. In this role, he develops applications for automated scoring of implicit motives using multi-model ensemble methodology. He is also the co-founder and CTO of AIMS, an AI-powered implicit motive assessment startup. Andrew’s presentation, with his advisor Joel Weinberger, PhD, “The ghost in the machine: Measuring implicit motives with AI,” at the 2024 American Psychological Association Annual Convention, was honored by Division 8 as a Top 5 Abstract.
The Research Project
Effective leadership is fundamentally driven by motivation—not what leaders say they want, but the unconscious needs that actually shape their behavior. For over seven decades, research has demonstrated that implicit motives—Achievement, Power, and Affiliation—predict leadership success more accurately than self-report measures, personality assessments, or cognitive ability tests (McClelland, Koestner, & Weinberger, 1989). These unconscious motivational patterns predict long-term career outcomes with approximately 67% accuracy, compared to 10-15% for traditional self-report instruments (Spangler, 1992).
The implications for leadership development are profound. Leaders with the “Leadership Motive Pattern”—high power motivation, low affiliation motivation, and high activity inhibition—are significantly more likely to succeed in management roles (Winter, 1991). The AT&T longitudinal studies demonstrated that two-thirds of managers with this profile were promoted, while two-thirds without it were not (Howard & Bray, 1988; Jacobs & McClelland, 1994). Power motivation predicts effective leadership; Achievement motivation predicts entrepreneurial success (Collins, Hanges, & Locke, 2004); and Affiliation motivation predicts teamwork and relationship-building.
Despite this robust evidence base, implicit motive assessment has remained largely inaccessible outside academic research. Traditional scoring of the Thematic Apperception Test (TAT) requires extensive expert training and 15-20 minutes per response (Smith, 1992). This labor intensity has prevented widespread adoption in leadership development, executive coaching, and organizational consulting—the very settings where it would provide the greatest value.
Relevance to t2pRI’s Mission
This research directly advances t2pRI’s mission to promote primary research in moral and purposeful leadership. Implicit motives represent the “why” behind leadership behavior—they reveal what genuinely drives a leader’s actions beyond conscious intention. By automating measurement of these fundamental motivational patterns, this project makes scientifically rigorous leadership assessment accessible at scale, enabling organizations to identify and develop leaders whose core motivations align with purposeful, values-driven leadership.
The research also connects to emotional intelligence by illuminating the unconscious processes that underlie interpersonal effectiveness. Power motivation shapes how leaders influence others; Affiliation motivation determines relationship-building capacity; Achievement motivation drives goal pursuit (McClelland & Burnham, 1976). These patterns interact to determine how leaders navigate organizational dynamics—core components of emotionally intelligent leadership.
Brief Summary of Relevant Literature
The study of implicit motives was systematically advanced by David McClelland at Harvard. McClelland’s foundational work (1987) established that Achievement, Power, and Affiliation motives are best assessed through narrative analysis because their unconscious nature makes them inaccessible to direct introspection. McClelland and Burnham’s classic Harvard Business Review article (1976) demonstrated that Power motivation predicts managerial effectiveness, while Spreier, Fontaine, and Malloy (2006) later showed the risks when achievement motivation becomes excessive.
Recent advances in artificial intelligence have created new possibilities for automating complex psychological assessments. Latif and Zhai (2023) demonstrated successful fine-tuning of large language models for automatic scoring applications. My preliminary research, presented at the APA Annual Convention (2024) and recognized by Division 8 as a top-five abstract, established that AI can score Achievement motivation with expert-level reliability (85% inter-rater agreement). In November 2025, our team presented updated findings at the Consortium for Research on Emotional Intelligence in Organizations (CREIO), demonstrating successful extension to Power and Affiliation motives using an advanced multi-model ensemble approach.
Research Question
This study addresses the following primary research question: Can a multi-model ensemble approach leveraging competitive reasoning models (including OpenAI, Claude, DeepSeek, and Gemini) score Power and Affiliation motives in TAT responses with accuracy equivalent to expert human raters (85% inter-rater reliability)? Secondary questions include: (1) What ensemble methodologies optimize performance for each motive category? (2) How does accuracy vary across stimulus images and response lengths? (3) What error patterns emerge to inform model refinement?