Do explanations make VQA models more predictable to a human?
Open accessThe DOI matches its Crossref record; 12 of the 12 references it cites were found in a registry or an index.
Abstract
A rich line of research attempts to make deep neural networks more transparent by generating human-interpretable 'explanations' of their decision process, especially for interactive tasks like Visual Question Answering (VQA).In this work, we analyze if existing explanations indeed make a VQA model -its responses as well as failures -more predictable to a human.Surprisingly, we find that they do not.On the other hand, we find that humanin-the-loop approaches that treat the model as a black-box do.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
“Why Should I Trust You?”: Explaining the Predictions of Any Classifier
Ribeiro, M., Singh, S., & Guestrin, C. (2016) · Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Demonstrations · doi:10.18653/v1/n16-3020 · Cited by 6,682
VQA: Visual Question Answering
Antol, S., Agrawal, A., Lu, J., Mitchell, M., Batra, D., Zitnick, C. L., & Parikh, D. (2015) · 2015 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2015.279 · Cited by 4,487
Self-citationMaking the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering
Goyal, Y., Khot, T., Summers-Stay, D., Batra, D., & Parikh, D. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.670 · Cited by 2,287
Self-citation(2013) · doi:10.1109/hri.2013.6483603 · Cited by 595 · cited 1×
“Dragan et al. (2013) describe the intent-expressiveness of a robot as its trajectory being expressive of its goal.”
BackgroundVQA: Visual Question Answering: www.visualqa.org
Agrawal, A., Lu, J., Antol, S., Mitchell, M., Zitnick, C. L., Parikh, D., & Batra, D. (2017) · International Journal of Computer Vision · doi:10.1007/s11263-016-0966-6 · Cited by 570 · cited 2×
“els of granularity (words, phrases, entire question) and at each level, has explicit attention mechanisms on both the image and the question1. It is trained on the train split of the VQA-1.0 dataset (Antol et al., 2015). Given an image and a question about the image, it outputs a probability distribution over 1000 ans…”
MethodInfluentialSelf-citationGenerating Visual Explanations
Hendricks, L. A., Akata, Z., Rohrbach, M., Donahue, J., Schiele, B., & Darrell, T. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46493-0_1 · Cited by 534
Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering
Goyal, Y., Khot, T., Agrawal, A., Summers-Stay, D., Batra, D., & Parikh, D. (2019) · International Journal of Computer Vision · doi:10.1007/s11263-018-1116-0 · Cited by 487 · cited 3×
“Additionally, the VQA 1.0 dataset contains label biases; therefore, the model is very likely to answer “white’ to a “what color” question (Goyal et al., 2016a).”
BackgroundInfluentialSelf-citationAnalyzing the Behavior of Visual Question Answering Models
Agrawal, A., Batra, D., & Parikh, D. (2016) · Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/d16-1203 · Cited by 297 · cited 2×
“Agrawal et al. (2016) outlines several such quirks.”
BackgroundSelf-citationTell me more?: the effects of mental model soundness on personalizing an intelligent agent
Kulesza, T., Stumpf, S., Burnett, M., & Kwan, I. (2012) · Proceedings of the SIGCHI Conference on Human Factors in Computing Systems · doi:10.1145/2207676.2207678 · Cited by 225 · cited 2×
“or helping them achieve an end goal (Narayanan et al., 2018; Kulesza et al., 2012).”
BackgroundNarayanan, M., Chen, E., He, J., Kim, B., Gershman, S., & Doshi-Velez, F. (2018) · arXiv · doi:10.48550/arxiv.1802.00682 · Cited by 153
Cited by
What is AI Literacy? Competencies and Design Considerations
Long, D., & Magerko, B. (2020) · Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3313831.3376727 · Cited by 3,241
From Recognition to Cognition: Visual Commonsense Reasoning
Zellers, R., Bisk, Y., Farhadi, A., & Choi, Y. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00688 · Cited by 769
“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].”
BackgroundDoes the Whole Exceed its Parts? The Effect of AI Explanations on Complementary Team Performance
Bansal, G., Wu, T., Zhou, J., Fok, R., Nushi, B., Kamar, E., Ribeiro, M. T., & Weld, D. (2021) · Proceedings of the 2021 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3411764.3445717 · Cited by 757
“As a result, various studies have evaluated the effect of explanations from different dimensions, including whether the explanation improves users’ trust on the AI [34, 68] or enables users to simulate the model predictions [13, 55], or assisting developers to debug and train a better ML model [7, 30].”
BackgroundDoes the Whole Exceed its Parts? The Effect of AI Explanations on Complementary Team Performance
Bansal, G., Wu, T., Zhou, J., Fok, R., Nushi, B., Kamar, E., Ribeiro, M. T., & Weld, D. S. (2020) · arXiv · doi:10.48550/arxiv.2006.14779 · Cited by 754
Lai, V., & Tan, C. (2019) · Proceedings of the Conference on Fairness, Accountability, and Transparency · doi:10.1145/3287560.3287590 · Cited by 377
“Concurrently, several recent studies have also examined how explanations relate to human performance [10, 23].”
BackgroundInterpreting CNNs via Decision Trees
Zhang, Q., Yang, Y., Ma, H., & Wu, Y. N. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00642 · Cited by 330
Towards Human-Centered Explainable AI: A Survey of User Studies for Model Explanations
Rong, Y., Leemann, T., Nguyen, T.-T., Fiedler, L., Qian, P., Unhelkar, V., Seidel, T., Kasneci, G., & Kasneci, E. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2023.3331846 · Cited by 278
Evaluating Explainable AI: Which Algorithmic Explanations Help Users Predict Model Behavior?
Hase, P., & Bansal, M. (2020) · Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics · doi:10.18653/v1/2020.acl-main.491 · Cited by 254
A Review on Explainability in Multimodal Deep Neural Nets
Joshi, G., Walambe, R., & Kotecha, K. (2021) · IEEE Access · doi:10.1109/access.2021.3070212 · Cited by 204
“Attention maps do not guarantee explanatory power [15] as the models and human perception does not concentrate on the same areas while providing the output.”
BackgroundAppropriate Reliance on AI Advice: Conceptualization and the Effect of Explanations
Schemmer, M., Kuehl, N., Benz, C., Bartos, A., & Satzger, G. (2023) · Proceedings of the 28th International Conference on Intelligent User Interfaces · doi:10.1145/3581641.3584066 · Cited by 208
“[12] analyze whether humans can learn to predict the AI’s behavior.”
BackgroundMeasuring Association Between Labels and Free-Text Rationales
Wiegreffe, S., Marasović, A., & Smith, N. A. (2021) · Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing · doi:10.18653/v1/2021.emnlp-main.804 · Cited by 119
“…has a rich history in machine learning inter-pretability as a reliable measure of rationale quality from the lens of utility to an end-user (Kim et al., 2016; Chandrasekaran et al., 2018; Hase and Bansal, 2020; Yeung et al., 2020; Poursabzi-Sangdeh et al., 2021; Rajagopal et al., 2021, i.a. ).”
BackgroundSmith-Renner, A., Fan, R., Birchfield, M., Wu, T., Boyd-Graber, J., Weld, D. S., & Findlater, L. (2020) · Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3313831.3376624 · Cited by 112
Measuring Association Between Labels and Free-Text Rationales
Wiegreffe, S., Marasović, A., & Smith, N. A. (2020) · arXiv · doi:10.48550/arxiv.2010.12762 · Cited by 106
A Meta-Analysis of the Utility of Explainable Artificial Intelligence in Human-AI Decision-Making
Schemmer, M., Hemmer, P., Nitsche, M., Kühl, N., & Vössing, M. (2022) · Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, and Society · doi:10.1145/3514094.3534128 · Cited by 91
“, whether humans are capable of predicting what a model would recommend (proxy tasks) [8, 13, 23] or whether explanations support them in model debugging [2, 32].”
BackgroundFok, R., & Weld, D. S. (2024) · AI Magazine · doi:10.1002/aaai.12182 · Cited by 81
Fok, R., & Weld, D. S. (2023) · arXiv · doi:10.48550/arxiv.2305.07722 · Cited by 54
Rajendran, S., Pan, W., Sabuncu, M. R., Chen, Y., Zhou, J., & Wang, F. (2024) · Patterns · doi:10.1016/j.patter.2023.100913 · Cited by 47
“87,110,111 These approaches individually focus on building interpretable components for either modality or modality interaction.”
BackgroundShould I Follow AI-based Advice? Measuring Appropriate Reliance in Human-AI Decision-Making
Schemmer, M., Hemmer, P., Kühl, N., Benz, C., & Satzger, G. (2022) · arXiv · doi:10.48550/arxiv.2204.06916 · Cited by 33
“For example, Chandrasekaran et al. [4] analyze whether humans can learn to predict the model behavior.”
BackgroundVisQA: X-raying Vision and Language Reasoning in Transformers
Jaunet, T., Kervadec, C., Vuillemot, R., Antipov, G., Baccouche, M., & Wolf, C. (2022) · IEEE Transactions on Visualization and Computer Graphics · doi:10.1109/tvcg.2021.3114683 · Cited by 30
Reasoning Models Don't Always Say What They Think
Chen, Y., Benton, J., Radhakrishnan, A., Uesato, J., Denison, C., Schulman, J., Somani, A., Hase, P., Wagner, M., Roger, F., et al. (2025) · arXiv · doi:10.48550/arxiv.2505.05410 · Cited by 18
“It measures whether humans can build precise and generalizable mental models of how the model processes various inputs through observing its CoTs (Doshi-Velez and Kim, 2017; Chandrasekaran et al., 2018; Hase and Bansal, 2020; Chen et al., 2024).”
Do Models Explain Themselves? Counterfactual Simulatability of Natural Language Explanations
Chen, Y., Zhong, R., Ri, N., Zhao, C., He, H., Steinhardt, J., Yu, Z., & McKeown, K. (2023) · arXiv · doi:10.48550/arxiv.2307.08678 · Cited by 5
“Simulatability measures how well humans can predict the model’s outputs based on its explanations (Doshi-Velez and Kim, 2017; Ribeiro et al., 2018; Chandrasekaran et al., 2018; Hase and Bansal, 2020); in particular, simulatability is a special case of faithfulness, which requires the output…”
BackgroundInfluentialMultiViz: Towards User-Centric Visualizations and Interpretations of Multimodal Models
Liang, P. P., Lyu, Y., Chhablani, G., Jain, N., Deng, Z., Wang, X., Morency, L.-P., & Salakhutdinov, R. (2023) · Extended Abstracts of the 2023 CHI Conference on Human Factors in Computing Systems · doi:10.1145/3544549.3585604 · Cited by 4
“Subsequently, their impact towards realworld applications has inspired recent research in visualizing and understanding their internal mechanics [15, 55, 68, 69, 71] as a step towards building interpretable, interactive, and reliable multimodal”
Background
Metadata
- Authors
- Chandrasekaran, A., Prabhu, V., Yadav, D., Chattopadhyay, P., & Parikh, D.
- Venue
- Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing
- Published
- 2018
- Pages
- 1036-1042
- OpenAlex
- W2899448251
- Cited by
- 55
- References
- 12
- FWCI
- 2.18
- Type
- conference paper
- Language
- en
- Open access
- hybrid
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.