Evaluating How Personalized AI Agents Influence Decision-Making, Self-Presentation, and Digital Identity Management: A Literature Review

Abstract

The increased interaction of humans with personalized AI agents has had a significant impact on how individuals engage with the digital world, influencing everything from the choices they make to the digital personas they present. While personalization is designed to optimize user experience, current scholarship remains fragmented, treating its influence on decision-making, self-presentation, and digital identity management as separate phenomena, thus failing to capture their systemic entanglement. This paper addresses this gap with a systematic literature review across Human-Computer Interaction, Communication Studies, and Psychology, building upon the theoretical concept of the extended self. The review reveals that personalized AI agents act as adaptive mediators, co-producing user behavior. Key findings confirm that personalized AI agents introduce systematic risks: algorithmic nudging erodes autonomy in decision-making; optimization for engagement leads to homogenization and loss of authenticity in self-presentation; and data-driven profiling actively constructs a constrained "Algorithmic Self" in digital identity management. Critically, these effects are found to operate through a continuous co-adaptive feedback loop. We propose an Integrative Evaluative Framework that formalizes this interconnected relationship as a Co-Adaptive Cycle. This framework offers a unified lens for scholarship. It generates critical implications for design, necessitating a shift toward systems that prioritize transparent identity modeling, support identity exploration, and ensure user ownership of the algorithmically extended self.