Quantifying Algorithmic Friction in Automated Resume Screening Systems

TL;DR

This paper introduces a method to measure algorithmic friction in automated resume screening, showing that semantic matching reduces false negatives compared to keyword-based methods, thereby improving hiring efficiency.

Contribution

It extends the Artificial Frictional Unemployment framework by providing an empirical measurement approach for algorithmic friction in resume screening systems.

Findings

Keyword-based screening has high algorithmic friction.

Semantic matching significantly reduces false negatives.

System-level analysis informs better recruitment system design.

Abstract

Automated resume screening systems are now a central part of hiring at scale, yet there is growing evidence that rigid screening logic can exclude qualified candidates before human review. In prior work, we introduced the concept of Artificial Frictional Unemployment to describe labor market inefficiencies arising from automated recruitment systems. This paper extends that framework by focusing on measurement. We present a method for quantifying algorithmic friction in resume screening pipelines by modeling screening as a classification task and defining friction as excess false negative rejection caused by semantic misinterpretation. Using controlled simulations, we compare deterministic keyword-based screening with vector-space semantic matching under identical qualification conditions. The results show that keyword-based screening exhibits high levels of algorithmic friction, while semantic representations substantially reduce false negative rejection without compromising precision. By treating algorithmic friction as a system-level property, this study provides an empirical basis for evaluating how recruitment system design affects matching efficiency in modern labor markets.