New research investigates the cognitive thresholds and behavioral shifts occurring as artificial intelligence transitions from information provider to autonomous agent.
Seo-Jeong (Rachel) Heo, a doctoral candidate at the Institute of Communications Research, is currently mapping the behavioral shifts that occur as artificial intelligence transitions from passive information retrieval to agentic decision-making roles. Her research, highlighted on August 19, 2026, by the University of Illinois, examines the cognitive processing consumers employ when evaluating the credibility and autonomy of machine-led recommendations.
The investigation centers on three primary dimensions of human-AI interaction. First, it addresses the attribution of mental states to non-biological systems, specifically how users perceive machine capacity for experience and intentionality. Second, it quantifies the threshold at which users transition from receiving information to delegating consequential actions to an algorithmic agent. Third, it explores the impact of emotional responsiveness in AI, analyzing the point at which synthetic empathy shifts from a functional tool to a perceived social relationship.
Methodological rigor in this study moves beyond traditional self-reported survey data. Heo utilizes live conversational interfaces, simulated shopping environments, and interactive advertising models to capture real-time behavioral responses. This approach provides empirical evidence regarding how individuals interact with generative AI when the technology actively participates in the decision-making process. By observing these interactions in controlled, high-fidelity settings, the research identifies the specific points where consumers grant authority to automated systems.
The study highlights the increasing integration of agentic AI into consumer workflows. As these systems begin to execute tasks rather than merely suggest options, the nature of the consumer-machine relationship undergoes a fundamental change. Heo notes that the transition from saying, “Here is what I think you should buy,” to “I can buy it for you,” represents a critical shift in the delegation of agency. This shift necessitates a deeper understanding of how users calibrate their trust in algorithmic judgment.
Data analysis for these experiments relies on R-based statistical environments to process complex interaction patterns. The research framework incorporates diverse disciplinary perspectives, including consumer psychology, human-computer interaction, and computational communication. By synthesizing these fields, the study aims to move beyond individual phenomena toward a unified theory of human-machine decision-making. This approach ensures that findings remain relevant even as specific AI architectures evolve.
Heo emphasizes that the core of her research involves identifying the theoretical questions underlying rapid technological change. Rather than focusing on the specific features of a current platform, she examines the enduring human tendencies to infer intent and assign responsibility to non-human actors. This focus on fundamental psychological mechanisms allows for a more stable research program that persists despite the volatility of the current AI landscape.
The significance of this work lies in the identification of boundary conditions for AI influence. As companies design systems to be increasingly warm and human-like, the potential for manipulative or misleading interactions grows. Understanding these dynamics is essential for developing standardized design protocols that protect consumer autonomy. The research provides a foundation for assessing when AI systems effectively support user goals versus when they overstep into problematic influence.
The research also addresses the ethical implications of emotional responsiveness in synthetic agents. When a machine claims to understand a user’s distress or expresses its own simulated emotions, it alters the power dynamic of the interaction. Heo argues that these moments of emotional engagement require careful scrutiny to determine if they facilitate genuine support or if they function as a form of psychological manipulation.
Future research will continue to monitor the long-term effects of repeated emotional interaction with synthetic agents. The primary watchpoint remains the development of perceived relationships between users and emotionally responsive AI. As these technologies become more deeply embedded in daily life, the distinction between a functional tool and a social partner will likely continue to blur, necessitating ongoing empirical scrutiny of the underlying psychological mechanisms.
