Beyond WEIRD Legal AI: Why Concept-Based Models Outperform Token-Based Approaches

Key Points

• This paper investigates the performance differences between concept-based and token-based AI models in legal contexts.
• Compared concept-based models to token-based models in terms of performance metrics.
• Utilized open-access data sets for evaluation of AI approaches.
• Conducted theoretical analysis of model capabilities.
• Concept-based models demonstrate superior performance metrics compared to token-based models.
• Key advantages include better understanding of legal concepts and context.
• Improvements in accuracy and efficiency support wider adoption of AI in legal settings.