This study originates from the observation that in the current paradigm shift of conversational AI
agents from 'functional tools' to 'social companions,' technological performance improvement
does not necessarily lead to user engagement. As AI perform...
This study originates from the observation that in the current paradigm shift of conversational AI
agents from 'functional tools' to 'social companions,' technological performance improvement
does not necessarily lead to user engagement. As AI performance becomes standardized, there is
a lack of practical guidelines for service designers on what key factors drive continuous user
engagement and what to prioritize.
Therefore, this study aimed to identify the relative importance and priority of multidimensional
factors affecting user engagement in 'character-based' conversational AI agents. Based on a review
of prior research (e.g., Svikhnushina & Pu, 2022; Pal et al., 2023; Chen et al., 2022), an AHP
(Analytic Hierarchy Process) model was designed with five Level 1 factors ('Emotional Bond,'
'Narrative Engagement,' 'Usability & Performance,' 'Trust & Safety,' 'Control & Transparency')
and 14 sub-factors. Valid responses were collected from 11 experts and quasi-experts in the
HCI/UX field for AHP analysis.
The analysis results (C.R. = 0.1158) showed that among the Level 1 factors, 'C4. Trust & Safety'
- 42 -
(46.6%) emerged as the top-priority factor, surpassing all others. This was followed by 'C3.
Usability & Performance' (19.8%) and 'C5. Control & Transparency' (18.8%). 'C1. Emotional
Bond' (8.5%) and 'C2. Narrative Engagement' (6.3%) had relatively lower importance.
In the comprehensive priority analysis of the 14 sub-factors, 'Trust & Safety' related factors ranked
highest: 1st 'Accuracy (Hallucination Suppression)' (23.1%), 2nd 'Harmful Content Suppression'
(11.9%), and 3rd 'Privacy Protection' (11.6%). This suggests that securing technical trust,
specifically accuracy and safety, is the fundamental prerequisite for the AI to function as a social
companion and establish an emotional bond with users.
Notably, in the relational aspect of AI, direct emotional expressions like 'Empathy' (13th, 1.8%)
and 'Intimacy' (14th, 1.3%) were ranked lowest. In contrast, factors that form the basis of a
persona's reliability, such as 'Character Consistency Maintenance' (4th, 10.8%) and 'Context
Awareness' (remembering the user) (7th, 5.4%), were valued much more highly. Furthermore,
'Personalization' (5th, 7.7%) was ranked higher than 'Response Speed' (6th, 6.8%) and 'Ease of
Use' (8th, 5.2%), confirming that customized experiences are a key competitive advantage in an
era of standardized performance.
This study empirically demonstrates that users' emotional satisfaction is achieved not merely
through direct emotional expressions, but only when (2) 'Consistency' (a non-breaking persona)
and (3) 'Context Awareness' (remembering past interactions) are combined upon the solid
foundation of (1) 'Trust' (ethical safety and accuracy). This provides strategic priorities for AI
service design and holds academic and practical significance for AI chatbot research as it shifts
from a 'performance-centric' to a 'relationship-centric' focus.