Paper Type

ERF

Abstract

AI systems with agentic capabilities have begun to reshape human-AI collaboration, yet it remains unclear how the structural design of such collaborations influences users' psychological experience and their evaluation of AI-generated outputs. This study investigates the locus of initiation and its effects on users' sense of agency and algorithmic appreciation. Drawing on agency theory, we propose that user-initiated interactions strengthen users' sense of agency, which in turn positively shapes their appreciation of AI-generated outputs. We further propose that user-based steerability moderates this relationship, such that the ability to direct and configure AI behaviour attenuates the psychological costs of system-initiated interactions. To test our model, we employ a 2 (Locus of Initiation: User-Initiated vs. System-Initiated) × 2 (User-Based Steerability: Low vs. High) between-subjects longitudinal experiment conducted over four weeks. In this experiment, the AI system serves as a writing assistant, generating analytical suggestions for a business consulting task. The findings will contribute to IS literature on human–AI collaboration, algorithmic appreciation, and interaction design, and offer practical guidance for developers on how structural design choices can preserve users' sense of agency as AI systems exhibit growing agentic capabilities.

Paper Number

1327

Comments

NEXTTRANS

Download

Share

COinS
 
Aug 15th, 12:00 AM

Agency Over Time: How Initiation and Steerability Shape User Experience with AI Systems Showing Agentic Capabilities

AI systems with agentic capabilities have begun to reshape human-AI collaboration, yet it remains unclear how the structural design of such collaborations influences users' psychological experience and their evaluation of AI-generated outputs. This study investigates the locus of initiation and its effects on users' sense of agency and algorithmic appreciation. Drawing on agency theory, we propose that user-initiated interactions strengthen users' sense of agency, which in turn positively shapes their appreciation of AI-generated outputs. We further propose that user-based steerability moderates this relationship, such that the ability to direct and configure AI behaviour attenuates the psychological costs of system-initiated interactions. To test our model, we employ a 2 (Locus of Initiation: User-Initiated vs. System-Initiated) × 2 (User-Based Steerability: Low vs. High) between-subjects longitudinal experiment conducted over four weeks. In this experiment, the AI system serves as a writing assistant, generating analytical suggestions for a business consulting task. The findings will contribute to IS literature on human–AI collaboration, algorithmic appreciation, and interaction design, and offer practical guidance for developers on how structural design choices can preserve users' sense of agency as AI systems exhibit growing agentic capabilities.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.