Zur Seitenansicht

Titelaufnahme

Links
Download
Klassifikation
Abstract
  • The article proposes simulating stakeholder workshops in product development using persona-based generative-AI agents. It argues that real workshops are hard to coordinate and that earlystage concept creation could benefit from a scalable, repeatable alternative. The approach instantiates multiple persona-driven agents representing distinct stakeholder roles (e.g., user, designer, safety engineer, and cost analyst) within an agent-based simulation framework (TinyTroupe). Agents are guided through a structured workshop workflow: setup, document review, idea generation, detailed recommendations, trade-off reasoning, and prioritization. During the process, agents can ask targeted questions to a human facilitator, and the answers can be stored in a shared knowledge base to improve consistency across rounds. A demonstration on an educational SCARA robot kit shows how to produce ranked improvement recommendations and synthesize different stakeholder perspectives into a consolidated outcome. The paper generalizes the workflow into a concept model and identifies key research gaps: generating empirically grounded personas, measuring how persona traits influence outcomes, maintaining persona consistency over long dialogues, validating outputs against human baselines, and scaling to richer product representations and tool access (e.g., structured product models and engineering artifacts). It also notes limited generalizability because the evaluation covers only one scenario and lacks systematic quality metrics.
Statistik
  • Das PDF-Dokument wurde 5 mal heruntergeladen.
Lizenz-/Rechtehinweis