The role of (voluntary) large language model use and individual needs in shaping work design

BackgroundLarge Language Models (LLMs) are increasingly adopted in organizational practice to support task execution. However, their impact on work design, and ultimately on employee outcomes, remains unclear.ObjectiveDrawing on theoretical frameworks regarding the effects of technologies on work design, we investigate potentially negative effects of LLM use on three work characteristics (decision-making autonomy, information processing, role overload) and on decision-making performance. We examine the mitigating role of voluntary LLM use and the needs for autonomy and cognition as moderating factors.MethodsWe conducted an online-experiment with 505 German workers from various industries, who performed two work tasks and were randomly assigned to one of three groups: no LLM use, instructed LLM use, or voluntary LLM use. Participants rated perceived work characteristics after task completion. Individual need strengths and decision-making performance were assessed.ResultsFindings support that LLM use reduces decision-making autonomy, information processing, and decision-making performance, but also decreases role overload. We found support for the mitigating role of voluntary LLM use and a moderating effect of need for autonomy, amplifying the effects on decision-making autonomy. Results did not support moderating effects for need for cognition.ConclusionsOverall, our experiment provides causal evidence that voluntary LLM use mitigates the negative effects on work characteristics, such as reduced decision-making autonomy and information processing, and losses in decision-making performance while reaping the benefits, such as reduced role overload. Voluntary use seems to be particularly important for workers with a high need for autonomy.