I-Corps: AI-assisted Job Search in the Aftermath of the Covid-19 Crisis

Regents Of The University Of Michigan - Ann Arbor, Ann Arbor MI

Investigators

Abstract

The broader impact/commercial potential of this I-Corps project is the development of a technology that will help customer-facing businesses hire from a pool of candidates that have no prior experience in the sector. The proposed technology may be used by hiring managers in mid- to large-scale employers of entry-level service workers. It is anticipated that the beachhead market will be retail. The key insights from preliminary customer discovery were that in large stores with >10k employees, the turnover in entry level positions is very high (e.g., 76% in retail in 2019) and each entry-level vacancy attracts a high volume of applications (e.g., up to 1,000 applications per position). Applicants for these vacancies are often indistinguishable given limited technical skills and previous experience. Therefore, current screening process for entry-level positions relies heavily on arbitrary filtering heuristics, individual judgement, and interviews, and, as a result, can be particularly prone to discrimination. In addition, employers aim to hire candidates with particular soft skills (e.g., conscientiousness, attention to detail, intrinsic motivation, communication), but find it difficult to objectively measure them. In addition, the cost of replacing a worker who leaves, excluding lost productivity, is high (typically 1.5 to 2.5 times the worker’s annual salary). By focusing on psychometric and simulated tasks, the proposed technology: a) reduces the risk of systemic discrimination while hiring; b) allows employers to cheaply, quickly and fairly identify best-suited candidates from a pool of indistinguishable candidates; c) have the ability to match workers to best-suited occupations by documenting the transferability of skills across occupations and sectors. The hypothesis is that the technology may reduce screening and hiring costs by more than 60% per head hired, in addition to finding employees that attrit at a lower rate. This I-Corps project is based on the development of machine learning classification algorithms that use psychometric profiles and performance on simulated tasks to identify best-suited candidates for entry-level customer facing occupations. These tasks have been designed in partnership with employers in each industry to mirror actual work. To select the optimal algorithm for each task, the proposed technology uses a cost function that takes into account the losses to the employer from wrongfully rejecting qualified individuals as well as from interviewing unqualified applicants. Additionally, the proposed technology uses labor market parameters in balancing false positive and false negative predictions in selecting and calibrating the algorithms, which pushes even the frontier in this space, and is beyond the capabilities of any practical solutions in the market today. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.