4.3.1. Direct Effect Analysis

Figure 2 displays the results of the SEM. The fit indices (χ2/df = 2.81, GFI = 0.921, NFI = 0.945, CFI = 0.958, RMSEA = 0.073) indicated the appropriateness of the structural model [83]. The path coefficients from perceived risk of COVID-19 to customers' attention (β = 0.422, *p* < 0.001), enthusiasm (β = 0.342, *p* < 0.001), and interaction (β = 0.358, *p* < 0.001) were positively significant. Therefore, H1a, H1b and H1c, which proposed that perceived risk of COVID-19 had positive influences on customer–robot engagement (respectively, attention, enthusiasm, and interaction), were supported.

**Figure 2.** Results of the SEM. \*\*\* *p* < 0.001.
