Towards Explainable Adjudicative Variance: Quantifying Judicial Discretion via Gated Multi-Task Learning

2026-06-25

AI Summary

Legal outcome prediction often conflates objective facts with judicial discretion, undermining model interpretability and fairness. This work proposes a judge-aware gated multi-task learning architecture that integrates a fine-grained legal taxonomy with a structured gating fusion mechanism to dynamically modulate reliance on judge identity, thereby enabling quantitative modeling of judicial discretion. The approach employs a LoRA-finetuned Gemma-4 encoder, judge embeddings, and structural regularization, achieving a new state-of-the-art on 13,937 UK Employment Tribunal decisions. It reduces parameter count by an order of magnitude compared to prompt-tuned large language models while significantly improving performance on the most challenging prediction categories and supporting interpretable localization of judicial discretion effects.

Abstract

Legal outcome prediction must disentangle objective case facts from adjudicative context. Merit-based rulings rely on factual evidence while technical disposals may hinge on judicial discretion. We propose a Judge-Aware Gated Multi-Task Learning architecture that explicitly models this distinction. We introduce a fine-grained outcome taxonomy to supervise the encoder, enforcing a structural regularization that disentangles distinct semantic pathways. This granular legal curriculum enables our Gated Fusion mechanism to dynamically modulate reliance on judge identity. We evaluate our approach on 13,937 UK Employment Tribunal decisions. We benchmark our design against supervised fine-tuning (SFT) of a Gemma-4 26B-A4B backbone, in which judge identity and the taxonomy are injected as prompt tokens or autoregressive output targets. The two contextual signals compose only weakly when forced through a single autoregressive channel. In contrast, coupling a LoRA-adapted Gemma-4 encoder with our gated architecture defines a new state of the art on this benchmark while requiring an order of magnitude fewer trainable parameters than the generative SFT baselines, with gains concentrated on the most ambiguous and rarest outcome classes. Beyond accuracy, the architecture is interpretable; learned judge embeddings and calibration profiles localize the cases where adjudicative context drives the prediction. These results indicate that, for identity-conditioned classification of legal outcomes, the choice of conditioning interface dominates scale: differentiable structured composition yields more accurate, more parameter-efficient models than prompt-based composition over a substantially larger backbone.

Problem

• judicial discretion
• legal outcome prediction
• adjudicative variance
• explainability
• judge identity

Innovation

• Gated Multi-Task Learning
• Judicial Discretion
• Explainable AI
• Legal Outcome Prediction
• Structured Composition