MAEA: Multimodal Attribution for Embodied AI
This work addresses the need for interpretability and trust in multimodal embodied AI policies, though it is incremental as it applies existing attribution methods to a specific domain.
The authors tackled the problem of understanding multimodal perception in embodied AI by developing MAEA, a framework to compute global attributions per modality for differentiable policies, and demonstrated its utility for analyzing failure scenarios, biases, and policy robustness on the ALFRED dataset.
Understanding multimodal perception for embodied AI is an open question because such inputs may contain highly complementary as well as redundant information for the task. A relevant direction for multimodal policies is understanding the global trends of each modality at the fusion layer. To this end, we disentangle the attributions for visual, language, and previous action inputs across different policies trained on the ALFRED dataset. Attribution analysis can be utilized to rank and group the failure scenarios, investigate modeling and dataset biases, and critically analyze multimodal EAI policies for robustness and user trust before deployment. We present MAEA, a framework to compute global attributions per modality of any differentiable policy. In addition, we show how attributions enable lower-level behavior analysis in EAI policies for language and visual attributions.