Work

Directed Recognizability in AI-Biographical Conversational Agents: Design and Evaluation in the Digital Afterlife Context

Year: 2026
Type: preprint
Abstract: Generative AI enables digital legacy systems that transform personal data into interactive conversational agents. Yet authenticity in such systems remains underspecified, especially from the perspective of living Data Donors whose data are used to construct the agent. This paper introduces directed recognizability as a design and evaluation lens for AI-biographical conversational agents. We present an AI Biographer that guides users through a structured biographical interview, generates a personalized agent, and supports post-interaction self-recognition evaluation. In an exploratory mixed-methods study with 15 participants, agents’ directed recognizability were evaluated positively overall. Value coherence and situational response patterns were recognized most reliably, while emotional resonance and humor congruence were more fragile. The findings suggest that authenticity should not be treated as complete replication, but as a bounded, interactional form of recognizability grounded in salient values, communicative patterns, and biographical context.
Language: English
Open Access status: green
OpenAlex ID: W7167581358
Cites:
Cited by:
Related to: 10