i-Algebra: Towards Interactive Interpretability of Deep Neural Networks
This addresses the need for more user-centric interpretability in deep neural networks, particularly for domains requiring interactive decision explanations, though it is incremental as it builds on existing interpretability concepts with a novel interactive approach.
The paper tackles the problem of static and non-interactive interpretability methods for deep neural networks by introducing i-Algebra, an interactive framework that allows users to build custom analysis tools through composable operators, demonstrating promising usability in tasks like inspecting adversarial inputs and resolving model inconsistency.
Providing explanations for deep neural networks (DNNs) is essential for their use in domains wherein the interpretability of decisions is a critical prerequisite. Despite the plethora of work on interpreting DNNs, most existing solutions offer interpretability in an ad hoc, one-shot, and static manner, without accounting for the perception, understanding, or response of end-users, resulting in their poor usability in practice. In this paper, we argue that DNN interpretability should be implemented as the interactions between users and models. We present i-Algebra, a first-of-its-kind interactive framework for interpreting DNNs. At its core is a library of atomic, composable operators, which explain model behaviors at varying input granularity, during different inference stages, and from distinct interpretation perspectives. Leveraging a declarative query language, users are enabled to build various analysis tools (e.g., "drill-down", "comparative", "what-if" analysis) via flexibly composing such operators. We prototype i-Algebra and conduct user studies in a set of representative analysis tasks, including inspecting adversarial inputs, resolving model inconsistency, and cleansing contaminated data, all demonstrating its promising usability.