40816
post-template-default,single,single-post,postid-40816,single-format-standard,stockholm-core-2.4,qodef-qi--no-touch,qi-addons-for-elementor-1.6.7,select-theme-ver-9.5,ajax_fade,page_not_loaded,,qode_menu_,wpb-js-composer js-comp-ver-7.9,vc_responsive,elementor-default,elementor-kit-38031
Title Image

Speed, Accuracy, and Risk: Nevada’s Use of Artificial Intelligence in Unemployment Claims Appeals

Speed, Accuracy, and Risk: Nevada’s Use of Artificial Intelligence in Unemployment Claims Appeals

At the height of the COVID-19 pandemic, 30 percent of Nevada’s workforce was estimated to be unemployed, leading to the backlog of 40,000 unemployment benefit appeals cases in 2023.[1] Though this backlog has whittled to 10,000 cases this summer,[2] and the unemployment rate decreased to 5.5 percent for the month of August 2024,[3] the state holds the second-highest unemployment rate in the country.[4]

To continue to diminish this case backlog, Nevada plans to use Google’s artificial intelligence system to assist with the unemployment benefits appeals process.[5] This will make Nevada the first state to use this technology to speed up the resolution of a backlog of appeals cases.[6]The system’s implementation is estimated to expedite the ruling process from three hours to five minutes.[7]

Nevada’s Department of Employment, Training, and Rehabilitation (DETR) will utilize Vertex AI Studio, a Google cloud service, for this purpose.[8] The system will be adopted only to reference DETR’s database information, to tailor the system, and to provide accurate results[9] under a retrieval-augmented generation (RAG) model.[10]

Under DETR’s contract with Google, which will cost roughly $1.38 million,[11] the Department will transfer the transcripts from unemployment hearings and rulings for Google’s systems to compare with previous cases.[12] Though the process is expected to be launched in the coming months,[13] an exact release timeline has not been published.[14]

A state employee will review any AI recommendation for approval or denial of unemployment benefits.[15]This seemingly aligns with the Biden Administration’s guidelines on agency use of AI by creating a “human in the loop” process.[16] The process seeks to curb biases or hallucinations that impact a case’s outcome.[17] The decision will then be approved or edited with feedback for DETR to investigate.[18]

Although there are benefits to the system’s ability to expedite the process– like providing Nevadans with thousands of dollars in benefits quickly[19]– scholars worry DETR employees might feel pressured to authorize AI decisions on appeal with haste.[20] This may further backlog the appeals if individuals object to erroneous rulings,[21] and may complicate the appeals process in civil court, as “district court cannot substitute its own judgment for the judgment of the appeal referee.”[22]

The pressure for rapid approval is further complicated by the accuracy of Google’s model itself; research suggests these models provide “incorrect or misleading answers to questions between 17 and 33 percent of the time and returned incomplete responses between 18 and 63 percent of the time.”[23] Integrating AI models may also make it difficult to pinpoint where and why a review error occurred.[24]

Privacy concerns are also amuck. Documents associated with unemployment hearings include tax information, social security numbers, and personal details about a claimant’s health, family, and finances.[25] A DETR spokesperson assured Google would be unable to access Nevadan’s data processed by the system for any other purpose.[26]

Though the benefits of speed must not be discounted, pioneer adopters of Artificial Intelligence must remain diligent on the impact this technology may have not only on legal appeals processes, but also on individual recipients who will have to deal with these hurdles in an already complicated, trying process.

Footnotes[+]

Dawn Edelman

Dawn Edelman is a second-year J.D. candidate at Fordham University School of Law and a staff member of the Intellectual Property, Media & Entertainment Law Journal. She holds a B.A. in Economics and Political Science from Tulane University.