Synthesizing Action Sequences for Modifying Model Decisions
Abstract
When a model makes a consequential decision, e.g., denying someone a loan, it needs to additionally generate actionable, realistic feedback on what the person can do to favorably change the decision. We cast this problem through the lens of program synthesis, in which our goal is to synthesize an optimal (realistically cheapest or simplest) sequence of actions that if a person executes successfully can change their classification. We present a novel and general approach that combines search-based program synthesis and test-time adversarial attacks to construct action sequences over a domain-specific set of actions. We demonstrate the effectiveness of our approach on a number of deep neural networks.
Cite
Text
Ramakrishnan et al. "Synthesizing Action Sequences for Modifying Model Decisions." AAAI Conference on Artificial Intelligence, 2020. doi:10.1609/AAAI.V34I04.5996Markdown
[Ramakrishnan et al. "Synthesizing Action Sequences for Modifying Model Decisions." AAAI Conference on Artificial Intelligence, 2020.](https://mlanthology.org/aaai/2020/ramakrishnan2020aaai-synthesizing/) doi:10.1609/AAAI.V34I04.5996BibTeX
@inproceedings{ramakrishnan2020aaai-synthesizing,
title = {{Synthesizing Action Sequences for Modifying Model Decisions}},
author = {Ramakrishnan, Goutham and Lee, Yun Chan and Albarghouthi, Aws},
booktitle = {AAAI Conference on Artificial Intelligence},
year = {2020},
pages = {5462-5469},
doi = {10.1609/AAAI.V34I04.5996},
url = {https://mlanthology.org/aaai/2020/ramakrishnan2020aaai-synthesizing/}
}