Does Familiarity Breed Concern? The Divergent Effects of LLM Use and Digital Literacy on Perceived Job Insecurity: Logistic Regression Evidence from Turkey

Özkan P., Süral Özer P., Köymen İ., Deveci Kocakoç İ.

7. International Applied Statistics Congress, İstanbul, Turkey, 11 - 13 May 2026, pp.1-8, (Full Text)

  • Publication Type: Conference Paper / Full Text
  • City: İstanbul
  • Country: Turkey
  • Page Numbers: pp.1-8
  • Dokuz Eylül University Affiliated: Yes

Abstract

This study examines the relationship between large language model (LLM) use, digital literacy, and perceived job insecurity among employed individuals in Türkiye. The study draws on data from the 2024 Turkey Income and Living Conditions Survey (TGSS) conducted by the ISAR Research Center, based on a probability-based multi-stage stratified cluster sample.

A binary logistic regression strategy was employed across three dimensions of perceived job insecurity: automation anxiety, short-term job loss risk, and re-employment expectation. General digital literacy was operationalized through a five-item usage frequency index, while LLM use was treated as a conceptually distinct predictor. Four model specifications were estimated per outcome, incorporating digital skill and LLM use alongside attitudinal AI variables and demographic controls. Sensitivity analyses assessed the stability of coefficients under sequential control addition, and interaction terms were tested to examine the moderating role of educational attainment.

The analyses revealed a significant negative association between general digital literacy and perceived job loss risk, and a significant positive association between LLM use frequency and the same outcome, a divergent pattern indicating that the two variables do not represent the same technological practice. The protective effect of digital literacy attenuated substantially once educational attainment was introduced, while the positive association between LLM use and risk perception remained comparatively robust under controls. The interaction between digital literacy and high educational attainment emerged as significant, suggesting that the risk-reducing effect of digital skill is conditional on broader human capital endowment.

These findings suggest that automation anxiety maps onto educational fault lines rather than digital exposure per se, and that direct contact with LLMs may transform abstract automation narratives into concrete substitutability awareness, a mechanism with implications for workforce training policy and organizational AI integration strategies.

Keywords: Perceived Job İnsecurity, Large Language Models, Digital Literacy, Automation Anxiety, Logistic Regression, Turkish Labor Market