Paper Number
ECIS2026-2452
Paper Type
CRP
Abstract
Agentic artificial intelligence (AI) systems represent a new generation of autonomous technologies capable of independent decision-making, multi-step planning, and self-directed action. Manus AI exemplifies this paradigm by autonomously executing complex tasks through goal decomposition and workflow automation. Understanding user satisfaction and trust in such systems is critical for sustainable human–AI interaction. This study develops a data-driven model linking system quality, information quality, interaction quality, enjoyment, transparency, and perceived intelligence to user satisfaction and trust. After data cleaning, 55,667 English-language reviews were analyzed using LDA to identify experiential themes, RoBERTa for polarity and subjectivity features, expert construct mapping, and machine learning regression techniques. Enjoyment has the strongest effect on satisfaction, followed by transparency and system quality. Satisfaction strongly predicts trust, with affective satisfaction exerting a greater influence than evaluative satisfaction. These findings support a dual-path model where cognitive and affective factors jointly shape trust. The study extends the information systems success framework to agentic AI.
Recommended Citation
Dwivedi, Yogesh K.; AI-Sharafi, Mohammed A.; Helal, Mohamed Youssef Ibrahim; and Alzaeemi, Shehab, "From User Experience To Trust: A Data-Driven Model Of Satisfaction And Trust Formation In Agentic Artificial Intelligence" (2026). ECIS 2026 Proceedings. 29.
https://aisel.aisnet.org/ecis2026/cog_hbis/cog_hbis/29
From User Experience To Trust: A Data-Driven Model Of Satisfaction And Trust Formation In Agentic Artificial Intelligence
Agentic artificial intelligence (AI) systems represent a new generation of autonomous technologies capable of independent decision-making, multi-step planning, and self-directed action. Manus AI exemplifies this paradigm by autonomously executing complex tasks through goal decomposition and workflow automation. Understanding user satisfaction and trust in such systems is critical for sustainable human–AI interaction. This study develops a data-driven model linking system quality, information quality, interaction quality, enjoyment, transparency, and perceived intelligence to user satisfaction and trust. After data cleaning, 55,667 English-language reviews were analyzed using LDA to identify experiential themes, RoBERTa for polarity and subjectivity features, expert construct mapping, and machine learning regression techniques. Enjoyment has the strongest effect on satisfaction, followed by transparency and system quality. Satisfaction strongly predicts trust, with affective satisfaction exerting a greater influence than evaluative satisfaction. These findings support a dual-path model where cognitive and affective factors jointly shape trust. The study extends the information systems success framework to agentic AI.