Asylum seekers are individuals who flee to other countries to find sanctuary from the persecution suffered within the borders of their home countries. The U.N. High Commissioner for Refugees estimated that by mid-2021 there were nearly 4.4 million individuals actively seeking asylum worldwide, and the most recent data available surprisingly suggest that the United States granted asylum to only 31,429 persons in 2020.

The asylum system that is with us today was created when Congress enacted the Refugee Act with the goal of “respond[ing] to the urgent needs of persons subject to persecution in their homelands” and “provid[ing] a permanent and systematic procedure for the admission to this country” for refugees and asylum seekers. Despite what may have been the best of intentions, courts and scholars today recognize that the U.S. asylum process “is in tatters.”

Although there are two methods by which an individual can gain asylum in the United States, this Comment principally concerns itself with affirmative asylum—the process by which a foreign national affirmatively applies to the U.S. Citizenship and Immigration Services (USCIS) for asylum. At the beginning of 2022, there were 196,714 affirmative asylum claims pending, and many applicants have waited in a state of legal limbo for over five years to receive a decision on their claim. To escape the indefinite queue, some have started bringing claims of unreasonable delay under the Administrative Procedure Act (APA) to federal courts.

Because there are groups of asylum seekers who may be especially harmed by multiyear delays in adjudication, this Comment undertakes two separate but related tasks. First, it assesses whether the avenue for relief available to advocates and asylum seekers—federal court litigation—is actually viable for its purported ends. This Comment concludes that it is not. Second, it proposes a novel agency-side adjudicative mechanism, implemented through artificial intelligence technology, to more adequately provide reliable relief to especially vulnerable asylum seekers. The proposal offers a sketch of the new mechanism, wrestles with how artificial intelligence may be incorporated into it, and finally explores how the transparency and accountability of the agency’s automated decision-making may still be attained through current administrative law doctrines.

寻求庇护者是逃往其他国家寻求避难所的人，他们在本国境内遭受迫害。联合国难民事务高级专员估计，到 2021 年年中，全球有近 440 万人在积极寻求庇护，而最新的数据令人惊讶地表明，美国在 2020 年仅向 31,429 人提供了庇护。

今天与我们同在的庇护制度是在国会颁布《难民法》时创建的，其目标是“回应在其家园受到迫害的人的紧急需求”和“提供永久和系统的程序难民和寻求庇护者进入这个国家”。尽管本意可能是最好的，但法院和学者今天承认美国的庇护程序“支离破碎”。

尽管个人可以通过两种方式在美国获得庇护，但本评论主要关注肯定性庇护——外国国民向美国公民及移民服务局 (USCIS) 肯定性地申请庇护的过程。到 2022 年初，有 196,714 项肯定的庇护申请未决，许多申请人在法律上处于悬而未决的状态等待了五年多，才收到对其申请的决定。为了摆脱无限期的排队，一些人开始根据《行政程序法》(APA) 向联邦法院提出不合理延误的索赔。

由于某些寻求庇护者群体可能会因多年的裁决​​延迟而受到特别伤害，因此本评论承担了两项独立但相关的任务。首先，它评估了倡导者和寻求庇护者可用的救济途径——联邦法院诉讼——是否真的可以达到其声称的目的。该评论的结论是事实并非如此。其次，它提出了一种新的代理端裁决机制，通过人工智能技术实施，以更充分地为特别脆弱的寻求庇护者提供可靠的救济。该提案提供了新机制的草图，探讨了如何将人工智能纳入其中，并最终探讨了如何通过当前的行政法学说仍然可以实现该机构自动决策的透明度和问责制。