• JUSTICE BY PROXY: WHEN AI BECOMES THE VOICE OF THE SILENT IN ASYLUM ADJUDICATION

This research explores the ability of advanced artificial intelligence (AI) tools to improve the persuasiveness, admissibility, and legal success of asylum applications with a focus on ensuing appeals decided upon by the European Court of Human Rights (ECHR). With the rise in global displacement and drastic shortages of legal representation for claimants, the research analyzes the capacity of state-of-the-art large language models (LLMs) — GPT, Perplexity, Grok, Claude, and Gemini — to algorithmically enhance stories of asylum seekers, thereby improving their admissibility and success rates. Employing a corpus of 621 ECHR asylum decisions, AI-fortified submissions were created and assessed with a three-prong approach: 1) quantitative text analysis, 2) predictive modeling involving a fine-tuned BERT-based legal judgment prediction (LJP) model, 3) semantic similarity analysis against prior winning instances. The results demonstrate statistically significant enhancements in text coherence, rhetorical power, and legal citing, with AI-advantaged representations consistently achieving greater predicted chances of winning. However, the low semantic similarity with past winning instances identifies emerging limitations of automatic narrative generation. Through its interdisciplinary synthesis of the humanities and formal sciences, including political science, linguistics, artificial intelligence, machine learning, and law, the study deems AI-facilitated enhancement a powerful promise for democratizing access to justice. Yet, equitable and accountable deployment is vital for protecting fairness and reliability in adjudication mechanisms for refugees.
Keywords: asylum adjudication, artificial intelligence, machine learning, access to justice