언어가 먼저 변한다: 직원 리뷰 텍스트 기반 조직 평판 하락 조기경보 시스템 개발 연구

초록

This study investigates whether linguistic changes in employee review texts can serve as early indicators of organizational reputation decline. Based on the premise that language shifts precede observable outcomes, the study constructs a company–month panel dataset using large-scale employee reviews from publicly listed Korean firms and develops an Early Warning System (EWS) that predicts future reputation decline using only contemporaneous linguistic and quantitative information. The analysis integrates natural language processing–based linguistic indicators with conventional quantitative metrics. The linguistic indicators capture psychological and cognitive characteristics embedded in review texts, including temporal orientation, emotional stance, and patterns of responsibility attribution, enabling early detection of shifts in employee expectations and perceptions. The results indicate that changes in linguistic patterns are systematically associated with subsequent organizational reputation decline. In particular, expressions of excessive future expectations, resignation-oriented language, and persistent external attribution of problems emerge as key early warning signals. These relationships remain robust after controlling for firm-specific characteristics, suggesting the broader applicability of language-based predictive models. The study contributes to organizational diagnosis and reputation research by demonstrating the value of linguistic signals as proactive indicators. Practically, it provides a methodological foundation for HR and organizational practitioners to use employee review texts as tools for early risk detection rather than solely as measures of satisfaction.

키워드