A Design Space for Explainable Ranking and Ranking Models

TL;DR

This paper introduces a comprehensive design space for explainable ranking models, aiming to improve transparency and user trust in multi-criteria decision-making systems by guiding future explainer development.

Contribution

It presents the first cross-domain design space for ranking explainers, characterizes existing methods and user groups, and offers a framework for designing targeted explanation solutions.

Findings

Characterized existing explainers using the design space

Identified three main user groups involved in ranking explanations

Provided a generative framework for future explainer development

Abstract

Item ranking systems support users in multi-criteria decision-making tasks. Users need to trust rankings and ranking algorithms to reflect user preferences nicely while avoiding systematic errors and biases. However, today only few approaches help end users, model developers, and analysts to explain rankings. We report on the study of explanation approaches from the perspectives of recommender systems, explainable AI, and visualization research and propose the first cross-domain design space for explainers of item rankings. In addition, we leverage the descriptive power of the design space to characterize a) existing explainers and b) three main user groups involved in ranking explanation tasks. The generative power of the design space is a means for future designers and developers to create more target-oriented solutions in this only weakly exploited space.