SkillUp NavGPT Evaluation

Overview

The SkillUp Coalition, launched in 2020, is a nonprofit organization dedicated to supporting young people and adults as they increase their skills or learn new skills to obtain employment in fields with good opportunities for them to build careers. SkillUp provides career navigation and connects workers—particularly those who have low incomes, those who do not have college degrees, and those who are skilled through alternative routes (STARs)—to job training and employment opportunities, with the ultimate goal of improving their career outcomes. Many of these workers, however, face barriers to learning new skills. To best serve the needs of all its clients, SkillUp is pilot testing a new feature of its online career navigation tool, called NavGPT, that uses generative artificial intelligence (AI) to provide personalized support and services.

Generative AI creates new text, images, and other data based on the information it is given. It offers organizations a novel way to provide services and potentially serve more people. In the case of SkillUp, generative AI has the potential to provide high-quality and customized career guidance services that may be able to reach more workers, improve their access to training and job opportunities, and ultimately increase their earnings.

This project is a partnership among SkillUp and several other organizations: Brighthive is developing the underlying hardware NavGPT will run on. AdeptID is developing the AI algorithm the tool uses. Burning Glass will write a white paper synthesizing findings about the tool and the broader generative AI landscape. And MDRC is leading an evaluation of the tool. The evaluation, supported by the Gates Foundation, will provide some of the first rigorous evidence about the ability of generative AI to enhance workforce development initiatives.

MDRC’s evaluation aims to rigorously test and refine the capabilities of NavGPT through a two-phase approach. Phase 1 is a formative assessment of NavGPT that focuses on users’ engagement with the tool and that gauges the quality and relevance of the text the tool generates. Phase 2 is a randomized experiment of an enhanced version of the tool (based on insights drawn from Phase 1 about how to optimize it), focused on assessing its effects on users’ outcomes, including engagement with NavGPT and placement into and completion of training courses. This work will lay the foundation for a potential longer-term and more comprehensive evaluation of the tool’s effects on users’ employment and earnings.