Do explanations make VQA models more predictable to a human?

Arjun Chandrasekaran, Viraj Prabhu, Deshraj Yadav, Prithvijit Chattopadhyay, Devi Parikh


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 human-in-the-loop approaches that treat the model as a black-box do.
Anthology ID:
D18-1128
Volume:
Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing
Month:
October-November
Year:
2018
Address:
Brussels, Belgium
Venue:
EMNLP
SIG:
SIGDAT
Publisher:
Association for Computational Linguistics
Note:
Pages:
1036–1042
Language:
URL:
https://aclanthology.org/D18-1128
DOI:
10.18653/v1/D18-1128
Bibkey:
Copy Citation:
PDF:
https://aclanthology.org/D18-1128.pdf
Attachment:
 D18-1128.Attachment.zip