[Paper Review] Evidence-based explanation to promote fairness in AI systems

Juliana Jansen Ferreira, Mateus de Souza Monteiro|arXiv (Cornell University)|Mar 3, 2020

Explainable Artificial Intelligence (XAI)13 references4 citations

TL;DR

This paper proposes an evidence-based explanation framework to enhance fairness in AI systems by enabling users to understand and justify AI-assisted decisions through traceable, context-aware narratives. By integrating fairness-sensitive design into decision storytelling, the approach improves transparency and accountability in high-stakes AI applications such as judicial or recommendation systems.

ABSTRACT

As Artificial Intelligence (AI) technology gets more intertwined with every system, people are using AI to make decisions on their everyday activities. In simple contexts, such as Netflix recommendations, or in more complex context like in judicial scenarios, AI is part of people's decisions. People make decisions and usually, they need to explain their decision to others or in some matter. It is particularly critical in contexts where human expertise is central to decision-making. In order to explain their decisions with AI support, people need to understand how AI is part of that decision. When considering the aspect of fairness, the role that AI has on a decision-making process becomes even more sensitive since it affects the fairness and the responsibility of those people making the ultimate decision. We have been exploring an evidence-based explanation design approach to 'tell the story of a decision'. In this position paper, we discuss our approach for AI systems using fairness sensitive cases in the literature.

Motivation & Objective

• To address the growing need for explainable and fair AI in high-stakes decision-making contexts such as judicial systems and recommendation engines.
• To investigate how AI systems can support users in justifying decisions by providing traceable, evidence-based narratives.
• To examine the role of fairness in AI explanations and how it affects user trust and responsibility in decision outcomes.
• To develop a design approach that frames AI decisions as 'stories' supported by verifiable evidence, particularly in fairness-sensitive scenarios.
• To contribute to responsible AI by aligning explanation mechanisms with ethical decision-making principles and human accountability.

Proposed method

• Adopting an evidence-based explanation design approach to structure AI decisions as coherent, justifiable narratives.
• Integrating fairness considerations into the explanation process by highlighting data, model behavior, and decision logic relevant to equity.
• Using contextual storytelling to connect AI outputs with human decision-making, emphasizing traceability and auditability.
• Applying principles from human-computer interaction (HCI) to ensure explanations are interpretable and actionable for end-users.
• Leveraging case studies from fairness-sensitive domains to ground the explanation framework in real-world decision contexts.
• Designing explanations that emphasize the 'why' behind decisions, particularly when fairness trade-offs are involved.

Experimental results

Research questions

• RQ1 How can AI systems provide explanations that support fairness in decision-making processes?
• RQ2 What design principles enable users to understand and justify AI-assisted decisions through evidence-based narratives?
• RQ3 How does integrating fairness into explanations affect user trust and accountability in AI systems?
• RQ4 In what ways can AI explanations be structured to reflect the decision-making journey in fairness-sensitive contexts?
• RQ5 What role does traceability of data and model behavior play in promoting responsible AI use?

Key findings

• The evidence-based explanation approach enables users to reconstruct the decision-making process with verifiable, context-specific information.
• Fairness-sensitive explanations improve user awareness of potential biases and ethical trade-offs in AI outputs.
• Narrative-based explanations enhance transparency by linking AI decisions to underlying data and logic, supporting accountability.
• The framework supports human-in-the-loop decision-making by making AI contributions traceable and justifiable.
• The approach is particularly effective in high-stakes domains where fairness and responsibility are paramount.
• User understanding and trust in AI decisions increase when explanations are grounded in evidence and fairness considerations.