Improving Current XAI Methods for Legal AI

Note: This article is just one of 60+ sections from our full report titled: The 2024 Legal AI Retrospective - Key Lessons from the Past Year. Please download the full report to check any citations.

Improving Current XAI Methods

Explanation: Sometimes an LLM's answer changes based on 'how' you ask them to explain their thinking. We need to learn, based on all the different ways of asking them to explain their thinking, which ones are most reliable.

Challenge:

Limitations of current attribution methods

Research Direction:

Develop aggregation techniques and focus on model robustness

Challenge:

Noisy results in synthesis-based explanations

Research Direction:

Develop quality metrics and apply across diverse domains