Timezone: »
As black box explanations are increasingly being employed to establish model credibility in high stakes settings, it is important to ensure that these explanations are accurate and reliable. However, prior work demonstrates that explanations generated by state-of-the-art techniques are inconsistent, unstable, provide very little insight into their correctness and reliability, and are computationally inefficient. In this paper, we address the aforementioned challenges by developing a novel Bayesian framework for generating local explanations along with their associated uncertainty. We instantiate this framework to obtain Bayesian versions of LIME and KernelSHAP which output credible intervals for the feature importances, capturing the associated uncertainty. The resulting explanations not only enable us to make concrete inferences about their quality (e.g., there is a 95\% chance that the feature importance lies within the given range), but are also highly consistent and stable. We carry out a detailed theoretical analysis that leverages the aforementioned uncertainty to estimate how many perturbations to sample, and how to sample for faster convergence. This work makes the first attempt at addressing several critical issues with popular explanation methods in one shot, thereby generating consistent, stable, and reliable explanations with guarantees in a computationally efficient manner. Experimental evaluation with multiple real world datasets and user studies demonstrate that the efficacy of the proposed framework.
Author Information
Dylan Slack (University of California Irvine)
Sophie Hilgard (Harvard University)
Sameer Singh (University of California, Irvine)
Hima Lakkaraju (Harvard)
More from the Same Authors
-
2021 : Towards the Unification and Robustness of Perturbation and Gradient Based Explanations »
· Sushant Agarwal · Shahin Jabbari · Chirag Agarwal · Sohini Upadhyay · Steven Wu · Hima Lakkaraju -
2021 : On the Connections between Counterfactual Explanations and Adversarial Examples »
· Martin Pawelczyk · Shalmali Joshi · Chirag Agarwal · Sohini Upadhyay · Hima Lakkaraju -
2021 : Towards a Rigorous Theoretical Analysis and Evaluation of GNN Explanations »
· Chirag Agarwal · Marinka Zitnik · Hima Lakkaraju -
2021 : What will it take to generate fairness-preserving explanations? »
· Jessica Dai · Sohini Upadhyay · Hima Lakkaraju -
2021 : Feature Attributions and Counterfactual Explanations Can Be Manipulated »
· Dylan Slack · Sophie Hilgard · Sameer Singh · Hima Lakkaraju -
2021 : On the Connections between Counterfactual Explanations and Adversarial Examples »
Martin Pawelczyk · Shalmali Joshi · Chirag Agarwal · Sohini Upadhyay · Hima Lakkaraju -
2021 : Towards Robust and Reliable Algorithmic Recourse »
Sohini Upadhyay · Shalmali Joshi · Hima Lakkaraju -
2021 : Towards a Unified Framework for Fair and Stable Graph Representation Learning »
Chirag Agarwal · Hima Lakkaraju · Marinka Zitnik -
2021 : Reliable Post hoc Explanations: Modeling Uncertainty in Explainability »
Dylan Slack · Sophie Hilgard · Sameer Singh · Hima Lakkaraju -
2022 : SAFER: Data-Efficient and Safe Reinforcement Learning via Skill Acquisition »
Dylan Slack · Yinlam Chow · Bo Dai · Nevan Wichers -
2023 Poster: Do Embodied Agents Dream of Pixelated Sheep: Embodied Decision Making using Language Guided World Modelling »
Kolby Nottingham · Prithviraj Ammanabrolu · Alane Suhr · Yejin Choi · Hannaneh Hajishirzi · Sameer Singh · Roy Fox -
2023 Tutorial: Responsible AI for Generative AI in Practice: Lessons Learned and Open Challenges »
Hima Lakkaraju · Krishnaram Kenthapadi · Nazneen Rajani -
2022 Workshop: New Frontiers in Adversarial Machine Learning »
Sijia Liu · Pin-Yu Chen · Dongxiao Zhu · Eric Wong · Kathrin Grosse · Hima Lakkaraju · Sanmi Koyejo -
2021 Workshop: ICML Workshop on Algorithmic Recourse »
Stratis Tsirtsis · Amir-Hossein Karimi · Ana Lucic · Manuel Gomez-Rodriguez · Isabel Valera · Hima Lakkaraju -
2021 : Towards Robust and Reliable Model Explanations for Healthcare »
Hima Lakkaraju -
2021 Poster: Calibrate Before Use: Improving Few-shot Performance of Language Models »
Tony Z. Zhao · Eric Wallace · Shi Feng · Dan Klein · Sameer Singh -
2021 Oral: Calibrate Before Use: Improving Few-shot Performance of Language Models »
Tony Z. Zhao · Eric Wallace · Shi Feng · Dan Klein · Sameer Singh -
2021 Poster: Learning Representations by Humans, for Humans »
Sophie Hilgard · Nir Rosenfeld · Mahzarin Banaji · Jack Cao · David Parkes -
2021 Spotlight: Learning Representations by Humans, for Humans »
Sophie Hilgard · Nir Rosenfeld · Mahzarin Banaji · Jack Cao · David Parkes -
2021 Poster: Towards the Unification and Robustness of Perturbation and Gradient Based Explanations »
Sushant Agarwal · Shahin Jabbari · Chirag Agarwal · Sohini Upadhyay · Steven Wu · Hima Lakkaraju -
2021 Spotlight: Towards the Unification and Robustness of Perturbation and Gradient Based Explanations »
Sushant Agarwal · Shahin Jabbari · Chirag Agarwal · Sohini Upadhyay · Steven Wu · Hima Lakkaraju -
2020 Poster: Robust and Stable Black Box Explanations »
Hima Lakkaraju · Nino Arsov · Osbert Bastani