Deconfounding Legal Judgment Prediction for European Court of Human Rights Cases Towards Better Alignment with Experts
This addresses the issue of improving AI alignment with expert reasoning in legal domains, though it is incremental as it builds on existing methods for deconfounding.
The paper tackled the problem of Legal Judgment Prediction systems being misled by legally irrelevant surface signals, and demonstrated that their deconfounded model, using adversarial training based on expert-identified confounders, aligns better with expert rationales than baselines.
This work demonstrates that Legal Judgement Prediction systems without expert-informed adjustments can be vulnerable to shallow, distracting surface signals that arise from corpus construction, case distribution, and confounding factors. To mitigate this, we use domain expertise to strategically identify statistically predictive but legally irrelevant information. We adopt adversarial training to prevent the system from relying on it. We evaluate our deconfounded models by employing interpretability techniques and comparing to expert annotations. Quantitative experiments and qualitative analysis show that our deconfounded model consistently aligns better with expert rationales than baselines trained for prediction only. We further contribute a set of reference expert annotations to the validation and testing partitions of an existing benchmark dataset of European Court of Human Rights cases.