Figure 2
The research framework diagram shows a dashed, rounded rectangle on the left labeled “Service chatbot’s attributes”. Inside it, five vertically arranged ovals read “Functionality”, “Trustworthiness”, “Efficiency”, “Human-likeness”, and “Reference to the service”, from top to bottom. Arrows from these ovals point to a central oval labeled “Perceived interactivity”. The path coefficients shown are negative 0.017 from “Functionality”, 0.250 triple asterisk from “Trustworthiness”, 0.204 double asterisk from “Efficiency”, 0.262 triple asterisk from “Human-likeness”, and 0.184 single asterisk from “Reference to the service”. Above, “Perceived interactivity” appears “R-squared equals 0.533”. An arrow labeled 0.470 triple asterisk extends from “Perceived interactivity” to an oval labeled “Brand experience”, with “R-squared equals 0.221” shown near it. Another arrow labeled 0.267 triple asterisk extends from “Perceived interactivity” to an oval labeled “Parasocial relationships”. A vertical arrow labeled 0.567 triple asterisk points downwards from “Brand experience” to “Parasocial relationships”, with “R-squared equals 0.538” shown near “Parasocial relationships”. At the bottom right, a dashed, rounded rectangle labeled “Control variables” contains four ovals labeled “Education”, “Gender”, “Age”, and “Chatbot’s use duration”, from left to right. Dashed arrows from these ovals point to “Parasocial relationships”, labeled 0.015 from “Education”, 0.001 from “Gender”, negative 0.041 from “Age”, and 0.037 from “Chatbot’s use duration”.PLS results for the proposed model. Note: *p < 0.05, **p < 0.01, ***p < 0.001. Responsiveness was removed to avoid construct overlap due to its high correlations with efficiency and service reference