ExplainabilityTransparencyTrustworthinesscs.HC
Effective Representation to Capture Collaboration Behaviors between Explainer and User
Arjun Akula, Song-Chun Zhu
January 10, 2022
arXiv.org
4 citations
1 influential
Key Takeaways
-
Natural language is a practical bridge between AI systems and users: The research proposes a framework allowing people to interact with explainable AI models using natural language rather than technical interfaces, making AI transparency more accessible to non-experts who need to understand AI decisions.
-
Collaboration between explainers and users matters for effective AI transparency: The paper emphasizes capturing "collaboration behaviors" between the AI system providing explanations and the user receiving them, suggesting that one-way explanations are less effective than interactive dialogue for genuine understanding.
-
Current XAI focus is incomplete: While recent efforts have concentrated on building explainable AI models that justify predictions, this research highlights a gap—there's less attention paid to how users actually interact with and benefit from these explanations in practice.
-
Generic frameworks can improve XAI accessibility: By developing a framework designed to work across different types of explainable AI systems (rather than custom solutions for each model), the research suggests a path toward making explanation capabilities more widely available and standardized.
Abstract
An explainable AI (XAI) model aims to provide transparency (in the form of justification, explanation, etc) for its predictions or actions made by it. Recently, there has been a lot of focus on building XAI models, especially to provide explanations for understanding and interpreting the predictions made by deep learning models. At UCLA, we propose a generic framework to interact with an XAI model in natural language.