Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PersonaCite: VoC-Grounded Interviewable Agentic Synthetic AI Personas for Verifiable User and Design Research

Mario Truss

TL;DR

PersonaCite addresses the core problem of unverifiable, prompt-based AI personas by grounding interactive agents in real voice-of-customer data retrieved in real time. The approach constrains LLM outputs to evidence, abstains when data are insufficient, and supplies response-level provenance, enabling verifiable exploration of user perspectives within design workflows. Through a formative evaluation with 14 industry experts, the study demonstrates benefits in rapid hypothesis testing and transparency while reframing validity as a design variable governed by provenance and documentation. The work introduces Persona Provenance Cards to standardize responsible documentation and positions grounded, interaction-time evidence as a complement to direct user research for responsible AI-assisted design.

Abstract

LLM-based and agent-based synthetic personas are increasingly used in design and product decision-making, yet prior work shows that prompt-based personas often produce persuasive but unverifiable responses that obscure their evidentiary basis. We present PersonaCite, an agentic system that reframes AI personas as evidence-bounded research instruments through retrieval-augmented interaction. Unlike prior approaches that rely on prompt-based roleplaying, PersonaCite retrieves actual voice-of-customer artifacts during each conversation turn, constrains responses to retrieved evidence, explicitly abstains when evidence is missing, and provides response-level source attribution. Through semi-structured interviews and deployment study with 14 industry experts, we identify preliminary findings on perceived benefits, validity concerns, and design tensions, and propose Persona Provenance Cards as a documentation pattern for responsible AI persona use in human-centered design workflows.

PersonaCite: VoC-Grounded Interviewable Agentic Synthetic AI Personas for Verifiable User and Design Research

TL;DR

PersonaCite addresses the core problem of unverifiable, prompt-based AI personas by grounding interactive agents in real voice-of-customer data retrieved in real time. The approach constrains LLM outputs to evidence, abstains when data are insufficient, and supplies response-level provenance, enabling verifiable exploration of user perspectives within design workflows. Through a formative evaluation with 14 industry experts, the study demonstrates benefits in rapid hypothesis testing and transparency while reframing validity as a design variable governed by provenance and documentation. The work introduces Persona Provenance Cards to standardize responsible documentation and positions grounded, interaction-time evidence as a complement to direct user research for responsible AI-assisted design.

Abstract

LLM-based and agent-based synthetic personas are increasingly used in design and product decision-making, yet prior work shows that prompt-based personas often produce persuasive but unverifiable responses that obscure their evidentiary basis. We present PersonaCite, an agentic system that reframes AI personas as evidence-bounded research instruments through retrieval-augmented interaction. Unlike prior approaches that rely on prompt-based roleplaying, PersonaCite retrieves actual voice-of-customer artifacts during each conversation turn, constrains responses to retrieved evidence, explicitly abstains when evidence is missing, and provides response-level source attribution. Through semi-structured interviews and deployment study with 14 industry experts, we identify preliminary findings on perceived benefits, validity concerns, and design tensions, and propose Persona Provenance Cards as a documentation pattern for responsible AI persona use in human-centered design workflows.
Paper Structure (7 sections, 1 figure, 1 table)

This paper contains 7 sections, 1 figure, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. System Design
  5. Findings & Discussion
  6. Conclusion
  7. Appendix

Figures (1)

  • Figure 1: PersonaCite architecture. VoC data from social media and other channels is imported and stored. Multimodal AI identifies topics and derives personas, storing personas alongside vectorized post data. Persona interaction and reaction simulation with response-level source attribution is enabled via data-grounded conversational AI, with explicit knowledge gap recognition when evidence is insufficient.