Overlooked Factors in Concept-Based Explanations: Dataset Choice, Concept Learnability, and Human Capability

Keyphrases

• Concept-based Explanations 100%
• Learnability 100%
• Human Capabilities 100%
• Concept Base 66%
• Interpretability Methods 66%
• Model Output 33%
• User Interface 33%
• New Probes 33%
• Human Studies 33%
• Deep Neural Network 33%
• Training Model 33%
• Semantic Concept 33%

Computer Science

• Interpretability 100%
• Neural Network Model 50%
• User Interfaces 50%
• Future Development 50%
• Deep Neural Network 50%
• Semantic Concept 50%

Social Sciences

• Neural Network 100%
• Generalizable 100%
• User Interface 100%