Multimodal Explanations: Justifying Decisions and Pointing to the Evidence

By Dong Huk Park, Lisa Anne Hendricks, Zeynep Akata, Anna Rohrbach, Bernt Schiele, Trevor Darrell, Marcus Rohrbach. 2018

482 citations recorded by the source corpus.

Deep models that are both effective and explainable are desirable in many settings; prior explainable models have been unimodal, offering either image-based visualization of attention weights or text-based generation of post-hoc justifications. We propose a multimodal approach to explanation, and argue that the two modalities provide complementary explanatory strengths. We collect two new datasets to define and evaluate this task, and propose a novel model which can provide joint textual rationale generation and attention visualization. Our datasets define visual and textual justifications of a classification decision for activity recognition tasks (ACT-X) and for visual question answering tasks (VQA-X). We quantitatively show that training with the textual explanations not only yields better textual justification models, but also better localizes the evidence that supports the decision. We also qualitatively show cases where visual explanation is more insightful than textual explanation, and vice versa, supporting our thesis that multimodal explanation models offer significant benefits over unimodal approaches.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Multimodal Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Multimodal Learning

Citations in the corpus

Reframing Human-AI Collaboration for Generating Free-Text Explanations (2021, methodology): “…paradigm for producing free-text explanations is via direct supervision, i.e., training an autoregressive, generative language model to predict human-authored explanations directly (Kim et al., 2018; Park et al., 2018; Ehsan et al., 2018; Narang et al., 2020; Wiegreffe et al., 2021, i.a. ).”
Few-Shot Out-of-Domain Transfer Learning of Natural Language Explanations (2021, background): “…labels at training time and generate such explanations for their decisions at deployment time (Hendricks et al., 2016; Ling et al., 2017; Park et al., 2018; Camburu et al., 2018; Kim et al., 2018; Rajani et al., 2019; Camburu et al., 2020; Narang et al., 2020; Kumar and Taluk-dar, 2020;…”
e-SNLI: Natural Language Inference with Natural Language Explanations (2018, positive): “In this direction, and very similar to our approach, is the recent work by Park et al. [23], who introduce two datasets of natural language explanations for the tasks of visual question-answering and activity recognition.”
An Empirical Study of GPT-3 for Few-Shot Knowledge-Based VQA (2021, methodology): “Alternatively, we perform answer rationale prediction (Li et al. 2018; Park et al. 2018; Zellers et al. 2019) in a zero-shot manner, and generate answer rationale as an open-ended text generation task.”
ViperGPT: Visual Inference via Python Execution for Reasoning (2023, background): “Many approaches provide explanations in the form of pixel importance, à la Grad-CAM [47, 65, 11, 41], some also providing textual explanations [41].”
From Recognition to Cognition: Visual Commonsense Reasoning (2018, background): “This has motivated work in having models provide explanations for their behavior, in the form of a natural language sentence [31, 9, 41] or an attention map [32, 35, 37].”
GQA: A New Dataset for Real-World Visual Reasoning and Compositional Question Answering (2019, neutral): “[26], and Li et al .”
Explaining Explanations: An Overview of Interpretability of Machine Learning (2018, background): “Datasets of human attention have been created [58], [59]; these allow systems to be evaluated according to how closely and their internal attention resembles human attention.”
Learning to Explain With Complemental Examples (2018, background): “As a different stream, some works trained the generative model that outputs explanations with natural language [12, 21] in a post-hoc manner.”
Explainable AI: The New 42? (2018, background): “Park et al. [21] propose a multimodal approach to explanation, and argue that the two modalities provide complementary explanatory strengths.”
Conservative Q-Improvement: Reinforcement Learning for an Interpretable Decision-Tree Policy (2019, background): “, we focus on learning a decision tree policy. Decision trees are frequently used to explain processes to humans, such as when a deep network is trained to explain itself [Alaniz and Akata, 2019; Huk Park et al., 2018]. Decision trees can be represented by text or graphically in a way that a human c”
Textual Explanations for Self-Driving Vehicles (2018, methodology): “[21] combine attention-based model and a textual justification system to produce an interpretable model.”
Grounded Video Description (2018, background): “While researchers have started to discover and study these problems for image description [14, 9, 24, 20],1 they are even more pronounced”
RISE: Randomized Input Sampling for Explanation of Black-box Models (2018, methodology): “To explain classiﬁcation decisions in images, previous works either visually ground image regions that strongly support the decision [18, 23] or generate a textual description of why the decision was made [10].”
e-ViL: A Dataset and Benchmark for Natural Language Explanations in Vision-Language Tasks (2021, methodology): “All the VL-NLE models M in this work consist of MT and ME modules, which can either be trained jointly or separately.”
Make Up Your Mind! Adversarial Generation of Inconsistent Natural Language Explanations (2019, background): “In order to increase users’ trust in these systems, a growing number of works [4, 26, 19, 14, 21] enhance neural networks with an explanation generation module that is jointly trained to produce natural language explanations for their ﬁnal decisions.”