JudgmentLens: Human-AI Sensemaking of Complex Legal Judgments

Judicial judgments are authoritative but notoriously difficult for non-experts to interpret due to dense legal language and complex, distributed relationships between facts, evidence, and rulings. While generative AI can summarize these documents, it often creates a new layer of interpretation that is difficult for users to verify. This paper investigates the sensemaking breakdowns non-experts face and introduces JudgmentLens, an AI-augmented system designed to scaffold the reading process.

Through a mixed-methods formative study, the authors identified four key sensemaking barriers: structural (reconstructing relationships), interpretive (understanding why decisions were made), verification (trusting AI-generated content), and action (translating findings into next steps). JudgmentLens addresses these by providing persistent, interactive representations—such as timelines, logic graphs, and evidence matrices—that are directly linked to the original source text.

In a counterbalanced within-subject evaluation with 16 participants, JudgmentLens significantly outperformed conventional PDF reading. Participants completed tasks 28.9% faster, reported lower cognitive workload, and expressed higher self-reported decision understanding. While rubric-scored comprehension was numerically higher with the system, the difference did not reach statistical significance, suggesting the primary benefit lies in reducing the effort and friction of the sensemaking process rather than replacing human judgment.

An exploratory probe comparing the system to a standard PDF+DeepSeek workflow highlighted a critical distinction in human-AI collaboration. While conversational AI can answer specific questions, it leaves the upstream work of question formulation, information integration, and source verification to the user. In contrast, JudgmentLens externalizes this upstream sensemaking, allowing users to navigate the case structure and verify interpretations against the source text more effectively.

This research demonstrates that for high-stakes documents, the value of AI lies not just in generating answers, but in providing an inspectable, source-grounded environment. By making provenance an interactive, user-facing feature, the system helps non-experts maintain epistemic agency—the ability to question and verify AI-generated interpretations—rather than passively accepting them. This design strategy offers a roadmap for building AI tools that support, rather than replace, human reasoning in complex, authoritative domains.