A 14-code content model for local post-hoc AI explanations, derived from 325 user statements and validated by experts with high reliability scores.
hub Canonical reference
Metrics for Explainable AI: Challenges and Prospects
Canonical reference. 80% of citing Pith papers cite this work as background.
15
Pith papers citing it
220
external citations · Pith
Background
80% of classified citations
abstract
The question addressed in this paper is: If we present to a user an AI system that explains how it works, how do we know whether the explanation works and the user has achieved a pragmatic understanding of the AI? In other words, how do we know that an explanainable AI system (XAI) is any good? Our focus is on the key concepts of measurement. We discuss specific methods for evaluating: (1) the goodness of explanations, (2) whether users are satisfied by explanations, (3) how well users understand the AI systems, (4) how curiosity motivates the search for explanations, (5) whether the user's trust and reliance on the AI are appropriate, and finally, (6) how the human-XAI work system performs. The recommendations we present derive from our integration of extensive research literatures and our own psychometric evaluations.
hub tools
citation-role summary
background 5
citation-polarity summary
roles
background 5representative citing papers
Interpretability research should be judged by actionability—the degree to which its insights support concrete decisions and interventions—rather than explanatory power alone.
Explanation preferences for AI privacy redaction vary systematically with domain and redaction amount; giving users their preferred styles raises trust over random or no explanations.
INSIGHTS creates manageable global summaries of time series model behavior by balancing sample importance and diversity with domain-specific utility functions, validated via experiments and user studies.
In interviews with 11 Portuguese-language model developers, four AI ethics tools guided general ethical reflection but failed to surface Portuguese-specific harms like cultural misrepresentation and low language performance.
The paper introduces knowledge affordances as declarative, relational descriptions of knowledge sources to guide information seeking in hybrid human-AI environments.
Standard LLM chats produce high perceived understanding but low objective learning in students, while future-self explanations best align confidence with actual gains and guided hints maximize learning with moderate workload.
Accuracy and understandability can be co-optimised for feature selection in tabular-data explanations while maintaining high classification performance.
Mod-Guide uses RAG with a community co-created corpus to make LLM moderation responses more contextually accurate for insensitive speech toward Bangladesh's Hindu and Chakma minorities, with mixed-method evaluation showing differences by ethnic background.
A position paper argues that post-hoc XAI explanations are unfaithful and paradoxical, proposing a shift to expert-based verification and certification of AI systems.
EZR combines active Naive Bayes sampling and decision-tree distillation to reach over 90% of best-known multi-objective optimization performance on 60 datasets while producing clearer explanations than LIME, SHAP or BreakDown.
AI explanations of slang improve non-native speakers' writing competence more than definitions or rewrites according to native raters, but users overestimate their skill and a performance gap with natives remains.
The paper calls for establishing explainable optimization (XOpt) as an interdisciplinary area to bridge the gap between optimization outputs and stakeholder needs for justification.
Chatbot AI systems often fail complex needs while projecting authority, contributing to deskilling, labor displacement, economic concentration, and high environmental costs, so alternative pluralistic and task-specific designs are needed.
citing papers explorer
-
What Should Explanations Contain? A Human-Centered Explanation Content Model for Local, Post-Hoc Explanations
A 14-code content model for local post-hoc AI explanations, derived from 325 user statements and validated by experts with high reliability scores.
-
Interpretability Can Be Actionable
Interpretability research should be judged by actionability—the degree to which its insights support concrete decisions and interventions—rather than explanatory power alone.
-
Exploring the Interaction of Explanation Styles, Context, and Trust of AI Privacy Redaction in AI-mediated Interactions
Explanation preferences for AI privacy redaction vary systematically with domain and redaction amount; giving users their preferred styles raises trust over random or no explanations.
-
INSIGHTS: Demonstration-Based Summaries of Time Series Predictors
INSIGHTS creates manageable global summaries of time series model behavior by balancing sample importance and diversity with domain-specific utility functions, validated via experiments and user studies.
-
Evaluation of AI Ethics Tools in Language Models: A Developers' Perspective Case Study
In interviews with 11 Portuguese-language model developers, four AI ethics tools guided general ethical reflection but failed to surface Portuguese-specific harms like cultural misrepresentation and low language performance.
-
Knowledge Affordances for Hybrid Human-AI Information Seeking
The paper introduces knowledge affordances as declarative, relational descriptions of knowledge sources to guide information seeking in hybrid human-AI environments.
-
Confidence Without Competence in AI-Assisted Knowledge Work
Standard LLM chats produce high perceived understanding but low objective learning in students, while future-self explanations best align confidence with actual gains and guided hints maximize learning with moderate workload.
-
Improving Explanations: Applying the Feature Understandability Scale for Cost-Sensitive Feature Selection
Accuracy and understandability can be co-optimised for feature selection in tabular-data explanations while maintaining high classification performance.
-
Mod-Guide: An LLM-based Content Moderation Feedback System to Address Insensitive Speech toward Indigenous Ethnic and Religious Minority Communities
Mod-Guide uses RAG with a community co-created corpus to make LLM moderation responses more contextually accurate for insensitive speech toward Bangladesh's Hindu and Chakma minorities, with mixed-method evaluation showing differences by ethnic background.
-
Beyond Explainable AI (XAI): An Overdue Paradigm Shift and Post-XAI Research Directions
A position paper argues that post-hoc XAI explanations are unfaithful and paradoxical, proposing a shift to expert-based verification and certification of AI systems.
-
Minimal Data, Maximum Clarity: A Heuristic for Explaining Optimization
EZR combines active Naive Bayes sampling and decision-tree distillation to reach over 90% of best-known multi-objective optimization performance on 60 datasets while producing clearer explanations than LIME, SHAP or BreakDown.
-
Reheat Nachos for Dinner? Evaluating AI Support for Cross-Cultural Communication of Neologisms
AI explanations of slang improve non-native speakers' writing competence more than definitions or rewrites according to native raters, but users overestimate their skill and a performance gap with natives remains.
-
Explainable Optimization: A Call for Interdisciplinary Action
The paper calls for establishing explainable optimization (XOpt) as an interdisciplinary area to bridge the gap between optimization outputs and stakeholder needs for justification.
-
What if AI systems weren't chatbots?
Chatbot AI systems often fail complex needs while projecting authority, contributing to deskilling, labor displacement, economic concentration, and high environmental costs, so alternative pluralistic and task-specific designs are needed.
- Supporting Calibrated Reliance in Human-AI Collaboration: Different Strategies for Different Tasks