[Paper Review] Enabling Value Sensitive AI Systems through Participatory Design Fictions

Q. Vera Liao, Michael Müller|arXiv (Cornell University)|Dec 13, 2019

Ethics and Social Impacts of AI22 references20 citations

TL;DR

This paper introduces participatory design fictions as a method to enable value-sensitive AI systems by engaging stakeholders in co-creating speculative narratives about AI use. By integrating participatory design with fictional scenarios, the approach elicits nuanced user values in context-specific domains, with a case study demonstrating its effectiveness in uncovering ethical concerns and stakeholder priorities in AI development.

ABSTRACT

Two general routes have been followed to develop artificial agents that are sensitive to human values---a top-down approach to encode values into the agents, and a bottom-up approach to learn from human actions, whether from real-world interactions or stories. Although both approaches have made exciting scientific progress, they may face challenges when applied to the current development practices of AI systems, which require the under-standing of the specific domains and specific stakeholders involved. In this work, we bring together perspectives from the human-computer interaction (HCI) community, where designing technologies sensitive to user values has been a longstanding focus. We highlight several well-established areas focusing on developing empirical methods for inquiring user values. Based on these methods, we propose participatory design fictions to study user values involved in AI systems and present preliminary results from a case study. With this paper, we invite the consideration of user-centered value inquiry and value learning.

Motivation & Objective

• To address the gap in current AI development practices that often overlook domain-specific values and stakeholder diversity.
• To overcome limitations of top-down value encoding and bottom-up value learning by grounding value inquiry in real-world contexts and stakeholder perspectives.
• To integrate insights from HCI’s long-standing focus on value-sensitive design into practical AI development workflows.
• To propose a novel method—participatory design fictions—that combines speculative storytelling with participatory design to elicit and explore user values.
• To demonstrate the feasibility and value of this method through a preliminary case study in a real-world domain.

Proposed method

• Develops participatory design fictions as a hybrid method merging participatory design with speculative fiction techniques.
• Engages diverse stakeholders in co-creating fictional narratives about future AI systems in specific application domains.
• Uses these fictional scenarios as tools to probe stakeholder values, ethical concerns, and contextual needs.
• Applies established empirical methods from HCI to analyze stakeholder responses and extract value-laden insights.
• Structures workshops around narrative prototypes to simulate AI interactions and elicit emotional, ethical, and practical responses.
• Synthesizes findings into design recommendations that inform value-sensitive AI system development.

Experimental results

Research questions

• RQ1 How can participatory design fictions effectively elicit stakeholder values in the context of AI system development?
• RQ2 What kinds of values emerge when stakeholders co-create fictional scenarios about future AI applications?
• RQ3 How do participatory design fictions compare to traditional value inquiry methods in capturing nuanced, context-specific values?
• RQ4 In what ways can fictional narratives support the identification of ethical trade-offs and design tensions in AI systems?
• RQ5 What role do stakeholder diversity and domain specificity play in shaping the values revealed through this method?

Key findings

• Participatory design fictions successfully elicited a broad range of values, including fairness, autonomy, transparency, and accountability, from diverse stakeholders.
• The method revealed context-specific ethical concerns not easily captured through traditional surveys or interviews.
• Stakeholders engaged more deeply with fictional narratives than with abstract value frameworks, indicating higher cognitive and emotional engagement.
• The case study demonstrated that fictional scenarios helped surface conflicting values, such as privacy versus utility, enabling early identification of design trade-offs.
• The approach enabled the identification of previously unconsidered stakeholder groups and their unique value priorities in AI system design.
• The resulting design insights were actionable and directly informed the specification of value-sensitive AI system requirements.