Explainability in AI: Emerging Trends and Techniques

The field of Artificial Intelligence (AI) is moving towards increased transparency and explainability, with a focus on developing techniques that can provide insights into the decision-making processes of AI models. Recent research has highlighted the importance of explainability in various applications, including language grounding, autonomous vehicles, and healthcare. The development of counterfactual explanations, model-agnostic approaches, and world models are some of the innovative techniques being explored to address the black box problem in AI. These techniques aim to provide meaningful explanations that can help practitioners and engineers improve model performance, identify biases, and enhance trust in AI systems.

Noteworthy papers in this area include:

  • Why Are You Wrong? Counterfactual Explanations for Language Grounding with 3D Objects, which presents a method for generating counterfactual examples to explain incorrect predictions.
  • Realistic Counterfactual Explanations for Machine Learning-Controlled Mobile Robots using 2D LiDAR, which introduces a novel approach for generating realistic counterfactual explanations in mobile robotics.
  • Explainable Reinforcement Learning Agents Using World Models, which proposes a technique for using world models to generate explanations for model-based deep RL agents.

Sources

Why Are You Wrong? Counterfactual Explanations for Language Grounding with 3D Objects

What Do People Want to Know About Artificial Intelligence (AI)? The Importance of Answering End-User Questions to Explain Autonomous Vehicle (AV) Decisions

Realistic Counterfactual Explanations for Machine Learning-Controlled Mobile Robots using 2D LiDAR

Explainable AI the Latest Advancements and New Trends

Explainable Artificial Intelligence Techniques for Software Development Lifecycle: A Phase-specific Survey

Accountability of Generative AI: Exploring a Precautionary Approach for "Artificially Created Nature"

Design Requirements for Patient-Centered Digital Health Applications: Supporting Patients' Values in Postoperative Delirium Prevention

A Theoretical Framework for Explaining Reinforcement Learning with Shapley Values

Explainable Reinforcement Learning Agents Using World Models

Explaining Autonomous Vehicles with Intention-aware Policy Graphs

PreCare: Designing AI Assistants for Advance Care Planning (ACP) to Enhance Personal Value Exploration, Patient Knowledge, and Decisional Confidence

Formalising Human-in-the-Loop: Computational Reductions, Failure Modes, and Legal-Moral Responsibility

Built with on top of