About

As AI models continue to advance in complexity and sophistication, understanding how they work and make decisions is becoming increasingly challenging. This challenge has prompted a surge of research into developing methods and tools that can enhance the transparency and explainability of these models. Nowadays, there are many such methods available, to the point that their specific applications have become somewhat unclear.

This workshop will specifically explore the diverse applications of explainable artificial intelligence (XAI) methods in various areas. The areas will include, but not limited to XAI in Healthcare, Natural Science, Auditing, Fairness, Natural Language Processing and Law. By examining the use of XAI in these fields, the workshop will provide attendees with insights into the latest trends and challenges within the different domains.

The workshop discussions aim to delve into the latest advancements in applied XAI and devise ways to further progress the field. The objective is to foster an open and productive dialogue that enhances our understanding of the potential opportunities and constraints of XAI and its impact across different domains. The purpose of this discourse is to identify strategies that can extend the frontiers of applied XAI and make notable progress in this rapidly evolving area. Specifically, the workshop aims to:

Topics covered

Examine various applications of XAI from the past and present
Discuss potential applications of XAI in the future
Identify the obstacles that hinder progress in each use case and how can we overcome them
Explore the necessary methodological requirements for applying XAI

Identify new domains where XAI can be useful in the future
Understand the inherent limitations of XAI
Explore whether insights gained from one use case can be transferred to other use cases

The workshop will provide a valuable learning opportunity for researchers, practitioners, and students seeking to apply XAI in their work, as it will feature presentations by experts in the field, as well as interactive discussions and insights into the latest trends and future directions in applied XAI. By bringing together a diverse group of participants with a shared interest in XAI, the workshop aims to foster collaboration, innovation, and knowledge sharing in this rapidly growing field.

Accepted Papers

A full list of accepted papers can be found here.

Oral Papers

Highlighted Reviewers

All Reviewers

Area Chairs

Dates

Note: all deadlines are in Anywhere on Earth (AoE).

Paper Submission

Submission deadline - ~~September 22~~ October 2 (23:59 AoE), 2023
Author notification - ~~October 20~~ October 27 (23:59 AoE), 2023
Camera ready deadline - November 22 (23:59 AoE), 2023

Workshop Event

Date: December 16, 2023

Schedule

Time	Event	Additional Information
8:50 - 9:00 AM	Opening Remarks
9:00 - 9:30 AM	IT1: Sameer Singh	Title: Explanations: Let's talk about them! Posthoc explanations aim to give end-users insights and understanding into the workings of complex machine learning models. Despite their potential, posthoc explanations have found limited use in real-world applications and, for some evaluation setups, fail to help end-users achieve their tasks effectively. In a survey we carried out with domain experts to understand why they do not use explanation techniques, they pointed out that explanations are static and inflexible, making it challenging to explore the model behavior intuitively. Based on these insights, we propose a shift towards natural language conversations as a promising avenue for future work for explainability: they are easy to use, flexible, and interactive. We introduce an initial version of such a system, TalkToModel, that uses LLMs to enable open-ended natural language conversations for machine learning explainability. In our evaluation, TalkToModel can accurately identify diverse user intents and support various user queries. Further, users strongly prefer TalkToModel over existing explainability systems, demonstrating the effectiveness of natural language interfaces in supporting model understanding. (This is work with Dylan Slack, Satya Krishna, Hima Lakkaraju, Chenhao Tan, and Yuxin Chen)
9:30 - 10:00 AM	IT2: Ulrike von Luxburg ※	Title: Theoretical guarantees for explainable AI? Explainable machine learning is often discussed as a tool to increase trust in machine learning systems. In my opinion, this can only work if the explanations are trustworthy themselves: we should be able to prove strong guarantees on the explanations provided. In my presentation I will argue that strong explanations in interesting scenarios might be difficult to achieve.
10:00 - 10:30 AM	Coffee Break & Interactive Games
10:30 - 11:00 AM	IT3: Su-In Lee	Title: Explainable AI, where we are and how to move forward for health AI.
11:00 - 12:00 PM	Panel Discussion	Moderator: Kamalika Chaudhuri. Panelists: Shai Ben-David, Julius Adebayo, Sameer Singh, Su-In Lee, Leilani Gilpin.
12:00 - 1:30 PM	Lunch
1:30 - 2:00 PM	IT4: Julius Adebayo	Title: Confronting the Faithfulness Challenge with Post-hoc Model Explanations. Explaining the output of a trained deep neural network has emerged as a key research challenge. Several classes of explanation methods (feature attributions, training point ranking, post-hoc concept attribution) have been proposed to address that challenge. However, despite significant research contributions, evidence points to their ineffectiveness. In this talk, I'll highlight a key challenge that undercuts the effectiveness of current post hoc explanations methods: faithfulness. A model's explanation is faithful if the feature importance score, induced by the explanation, indicates the magnitude of the change in the model's output, when that feature is ablated. However, consistent evidence indicates that post hoc explanations of large-scale deep nets, under standard training regimes, are unfaithful. I'll close with two vignettes: the first on emerging recipes for overcoming the faithfulness challenge, and the second on an alternative paradigm that involves developing intrinsically interpretable models.
2:00 - 3:00 PM	Poster Session 1	Paper ID: 1 - "Geometric Remove-and-Retrain (GOAR): Coordinate-Invariant eXplainable AI Assessment" Paper ID: 2 - "How Well Do Feature-Additive Explainers Explain Feature-Additive Predictors?" Paper ID: 4 - "Scale Alone Does not Improve Mechanistic Interpretability in Vision Models" Paper ID: 5 (DEMO) - "Utilizing Explainability Techniques for Reinforcement Learning Model Assurance" Paper ID: 8 - "GInX-Eval: Towards In-Distribution Evaluation of Graph Neural Network Explanations" Paper ID: 9 - "Towards the next generation explainable AI that promotes AI-human mutual understanding" Paper ID: 10 - "Assessment of the Reliablity of a Model's Decision by Generalizing Attribution to the Wavelet Domain" Paper ID: 11 - "Influence Based Approaches to Algorithmic Fairness: A Closer Look" Paper ID: 12 (DEMO) - "ExpLIMEable: An exploratory framework for LIME" Paper ID: 13 - "Understanding Scalable Perovskite Solar Cell Manufacturing with Explainable AI" Paper ID: 14 - "Detecting Spurious Correlations via Robust Visual Concepts in Real and AI-Generated Image Classification" Paper ID: 16 - "Explaining black box text modules in natural language with language models" Paper ID: 18 - "Lessons from Usable ML Deployments and Application to Wind Turbine Monitoring" Paper ID: 21 - "Beyond Concept Bottleneck Models: How to Make Black Boxes Intervenable?" Paper ID: 24 - "Emergence of Segmentation with Minimalistic White-Box Transformers" Paper ID: 25 - "Optimising Human-AI Collaboration by Learning Convincing Explanations" Paper ID: 26 - "Inherent Inconsistencies of Feature Importance" Paper ID: 27 - "Use Perturbations when Learning from Explanations" Paper ID: 28 - "Estimation of Concept Explanations Should be Uncertainty Aware" Paper ID: 29 - "Are VideoQA Models Truly Multimodal?" Paper ID: 30 - "Extracting human interpretable structure-property relationships in chemistry using XAI and large language models" Paper ID: 31 - "Towards Explanatory Model Monitoring" Paper ID: 32 - "Diagnosing Transformers: Illuminating Feature Spaces for Clinical Decision-Making" Paper ID: 33 - "Explainable AI in Music Performance: Case Studies from Live Coding and Sound Spatialisation" Paper ID: 34 - "COMET: Neural Cost Model Explanation Framework" Paper ID: 35 - "Piecewise Linear Parametrization of Policies: Towards Interpretable Deep Reinforcement Learning" Paper ID: 36 (DEMO) - "DeepDecipher: Accessing and Investigating Neuron Activation in Large Language Models" Paper ID: 37 (DEMO) - "Interactive Visual Feature Search" Paper ID: 40 - "AttributionLab: Faithfulness of Feature Attribution Under Controllable Environments" Paper ID: 42 - "FRUNI and FTREE synthetic knowledge graphs for evaluating explainability"
3:00 - 3:30 PM	Coffee Break & Interactive Games
3:30 - 4:00 PM	IT5: Leilani Gilpin	Title: Explaining Self-Driving Cars for Accountable Autonomy. Autonomous systems are prone to errors and failures without knowing why. In critical domains like driving, these autonomous counterparts must be able to recount their actions for safety, accountability, and trust. An explanation: a model-dependent reason or justification for the decision of the autonomous agent being assessed, is a key component for post-mortem failure analysis, but also for pre-deployment verification. I will present neuro-symbolic systems that use neural networks and commonsense knowledge to detect and explain unreasonable vehicle scenarios, even if the autonomous vehicle has not seen that error before. In the second part of the talk, I will motivate the use of explanations as a testing framework for autonomous systems. I will conclude by discussing new challenges at the intersection of explainable AI and autonomy toward autonomous vehicles systems that are explainable by design.
4:00 - 4:30 PM	Contributed Talks
4:30 - 5:30 PM	Poster Session 2	Paper ID: 44 - "GLANCE: Global to Local Architecture-Neutral Concept-based Explanations" Paper ID: 45 - "ReLax: An Efficient and Scalable Recourse Explanation Benchmarking Library using JAX" Paper ID: 46 - "Caution to the Exemplars: On the Intriguing Effects of Example Choice on Human Trust in XAI" Paper ID: 47 - "A Simple Scoring Function to Fool SHAP: Stealing from the One Above" Paper ID: 48 (DEMO) - "Transparent Anomaly Detection via Concept-based Explanations" Paper ID: 49 - "The Disagreement Problem in Faithfulness Metrics" Paper ID: 50 - "Diffusion-Guided Counterfactual Generation for Model Explainability" Paper ID: 52 - "Exploring Practitioner Perspectives On Training Data Attribution Explanations" Paper ID: 53 - "On Evaluating Explanation Utility for Human-AI Decision-Making in NLP" Paper ID: 54 (DEMO) - "Empowering Domain Experts to Detect Social Bias in Generative AI with User-Friendly Interfaces" Paper ID: 55 - "A Critical Survey on Fairness Benefits of XAI" Paper ID: 56 - "Stability Guarantees for Feature Attributions with Multiplicative Smoothing" Paper ID: 57 - "On the Consistency of GNN Explainability Methods" Paper ID: 59 - "Explaining Tree Model Decisions in Natural Language for Network Intrusion Detection" Paper ID: 63 - "Do Concept Bottleneck Models Obey Locality?" Paper ID: 64 - "Sanity Checks Revisited: An Exploration to Repair the Model Parameter Randomisation Test" Paper ID: 68 - "Visual Topics via Visual Vocabularies" Paper ID: 70 - "Explainable Reinforcement Learning for Alzheimer’s Disease Progression Prediction." Paper ID: 71 - "Explaining high-dimensional text classifiers" Paper ID: 72 - "Sum-of-Parts Models: Faithful Attributions for Groups of Features" Paper ID: 74 - "Explaining Longitudinal Clinical Outcomes using Domain-Knowledge driven Intermediate Concepts" Paper ID: 75 - "Rectifying Group Irregularities in Explanations for Distribution Shift" Paper ID: 76 - "Are Large Language Models Post Hoc Explainers?" Paper ID: 78 - "ObEy: Quantifiable Object-based Explainability without Ground-Truth Annotations" Paper ID: 79 - "Cost-aware counterfactuals for black box explanations" Paper ID: 83 - "Robust Recourse for Binary Allocation Problems" Paper ID: 84 (DEMO) - "Policy graphs in action: explaining single- and multi-agent behaviour using predicates" Paper ID: 85 - "On the Relationship Between Explanation and Prediction: A Causal View" Paper ID: 86 - "Interactive Model Correction with Natural Language"

※ Indicates virtual participation

All times are in Central Standard Time

Speakers

Julius Adebayo

Postdoctoral Fellow
Prescient Design

Leilani Gilpin

Assistant Professor
UC Santa Cruz

Su-In Lee

Professor
Paul G. Allen School of Computer Science & Engineering

Ulrike von Luxburg

Professor
University of Tübingen

Sameer Singh

Associate Professor
University of California, Irvine

Organisers

Chhavi Yadav

PhD student at UCSD, her interests lie in XAI, Secure Verification, Auditing and societal impacts of deep generative models

Michal Moshkovitz

Machine Learning Research Scientist at Bosch Research, she has been focused on developing the foundations of explainable machine learning

Bingqing Chen

Machine Learning Research Scientist at Bosch Research, her research lies at the intersection of machine learning and energy systems

Nave Frost

Research Scientist at eBay Research, his research interests focus on supplying explanations for data science applications

Suraj Srinivas

Postdoctoral research fellow at Harvard University, his research focuses on developing the foundations for interpretable machine learning

Valentyn Boreiko

PhD student at the University of Tübingen and a sabbatical student at Bosch Research, his research focuses on development of interpretability technique for vision classifiers

Hima Lakkaraju

Assistant Professor at Harvard University who focuses on the algorithmic and applied aspects of explainability, fairness, robustness, and privacy of machine learning models

Zico Kolter

Aassociate Professor at CMU, and chief scientist at Bosch Research, his work spans the intersection of machine learning and optimization

Dotan Di Castro

Research scientist and lab manager at Bosch Research, his research focuses on Reinforcement Learning and Computer Vision

Kamalika Chaudhuri

Associate Professor at UCSD and a Research Scientist at Meta AI, her research interests lie in the foundations of trustworthy machine learning

Contact information

Email: appliedXAI.neurips2023 [AT] gmail.com
Twitter: @XAI_in_Action

XAI in Action: Past, Present, and Future Applications