Paper Type
Complete
Paper Number
PACIS2026-1361
Description
The continued use of generative AI tools such as ChatGPT remains under-theorized despite rapid adoption. This study investigates how personal cognitions (self-efficacy, outcome expectations), social learning, and environmental cues jointly shape AI user satisfaction and continued use intention, and how task complexity and social support condition the satisfaction-continuance relationship. Survey data from 336 ChatGPT users in Taiwan are analyzed using PLS-SEM and triangulated with fuzzy-set qualitative comparative analysis (fsQCA). Results show that all four SCT antecedents significantly predict satisfaction, with environmental influence as the strongest predictor; satisfaction strongly predicts continuance; social support directly increases continuance but negatively moderates the satisfaction-continuance link, indicating a substitution dynamic; and task complexity exerts no direct effect but positively moderates the link, supporting a complexity-amplification logic. The fsQCA reveals three sufficient configurations and confirms causal asymmetry. The study contributes a dual-contingency framework and a methodological template integrating net-effect and configurational analyses for AI continuance research.
Recommended Citation
Wu, Chih-Lun, "The Impact of Task Complexity and Social Support on Continued ChatGPT Use Intention: A Social Cognitive Theory Perspective" (2026). PACIS 2026 Proceedings. 5.
https://aisel.aisnet.org/pacis2026/ai_ethic/ai_ethic/5
The Impact of Task Complexity and Social Support on Continued ChatGPT Use Intention: A Social Cognitive Theory Perspective
The continued use of generative AI tools such as ChatGPT remains under-theorized despite rapid adoption. This study investigates how personal cognitions (self-efficacy, outcome expectations), social learning, and environmental cues jointly shape AI user satisfaction and continued use intention, and how task complexity and social support condition the satisfaction-continuance relationship. Survey data from 336 ChatGPT users in Taiwan are analyzed using PLS-SEM and triangulated with fuzzy-set qualitative comparative analysis (fsQCA). Results show that all four SCT antecedents significantly predict satisfaction, with environmental influence as the strongest predictor; satisfaction strongly predicts continuance; social support directly increases continuance but negatively moderates the satisfaction-continuance link, indicating a substitution dynamic; and task complexity exerts no direct effect but positively moderates the link, supporting a complexity-amplification logic. The fsQCA reveals three sufficient configurations and confirms causal asymmetry. The study contributes a dual-contingency framework and a methodological template integrating net-effect and configurational analyses for AI continuance research.
Comments
03-EthicsSocietalImpact