4.3.2. Mediation Analysis

The bootstrapping procedure suggested by Hayes [77], with a confidence level of 95% and a bootstrap sample of 5000, was conducted to examine the mediating effect of social distancing. The analysis results are shown in Table 5. All the concerned 95% confidence intervals excluded the value of 0, thereby supporting the indirect effects of perceived risk on customers' attention (effect size = 0.088, SE = 0.026, 95% CI [0.045, 0.147]), enthusiasm (effect size = 0.078, SE = 0.021, 95% CI [0.043, 0.126]) and interaction (effect size = 0.099, SE = 0.024, 95% CI [0.059, 0.151]) through social distancing. These results implied social distancing mediated the effect of perceived risk on customer–robot engagement. Therefore, H2a, H2b, and H2c were supported.

**Table 5.** Mediating effect analysis results (*n* = 589).


Note. LL = Lower limit, UL = Upper limit, CI = Confidence interval. SE, standardized error. The value of the lower limit and that of the upper limit constitutes a confidence interval.

#### 4.3.3. Moderated Mediation Analysis

The bootstrapping procedure based on PROCESS macro suggested by Hayes [77], with a confidence level of 95% and a bootstrap sample of 5000, was conducted to examine H3a to H4c. The analysis results are shown in Table 6.


**Table 6.** Analysis results for the moderated mediation effect (*n* = 589).

Notes. DVs, dependent variables; SE, standardized error. Perceived risk as the independent variable, social distancing as the mediator, risk attitude, and health consciousness as moderators. Confidence interval (CI) was 95%. Bootstrap samples was 5000. Risk attitude: seeking = 0, avoiding = 1.

> Using attitude towards risk as the moderator, the index of moderated mediation was significant for customers' attention (index = 0.052, SE = 0.020, 95% CI [0.018, 0.098]), enthusiasm (index = 0.046, SE = 0.016, 95% CI [0.016, 0.085]), and interaction (index = 0.059, SE = 0.022, 95% CI [0.019, 0.104]), indicating the risk attitude moderated the mediating effects of social distancing on the relationship between perceived risk and customers' attention, enthusiasm, and interaction. For risk-averse consumers, social distancing significantly mediated the effect of perceived risk on customers' attention (effect size = 0.113, SE = 0.032, 95% CI [0.060, 0.182]), enthusiasm (effect size = 0.100, SE = 0.025, 95% CI [0.057, 0.156]), and interaction (effect size = 0.127, SE = 0.029, 95% CI [0.078, 0.189]). In contrast, for risk-seeking customers, the mediating effect of social distancing on customers' attention (effect size = 0.061, SE = 0.026, 95% CI [0.022, 0.125]), enthusiasm (effect size = 0.054, SE = 0.032, 95% CI [0.060, 0.182]), and interaction (effect size = 0.069, SE = 0.025, 95% CI [0.029, 0.128]) were still significant but the effect sizes were considerably reduced (attention: from 0.113 to 0.061; enthusiasm: from 0.100 to 0.054; interaction: from 0.127 to 0.069, Figure 3A), in support of H3a, H3b, and H3c.

**Figure 3.** Conditional indirect effect. (**A**) Moderating role of risk attitude. (**B**) Moderating role of health consciousness.

Using health consciousness as the moderator, the index of moderated mediation was not significant for customers' attention (index = 0.006, SE = 0.010, 95% CI [−0.013, 0.025]), enthusiasm (index = 0.005, SE = 0.008, 95% CI [−0.012, 0.021]), and interaction (index = 0.007, SE = 0.011, 95% CI [−0.015, 0.027]), which means the mediating effect sizes were not significant difference between high and low levels of health consciousness (Figure 3B). Specifically, for high levels of health consciousness, social distancing significantly mediated the effect of perceived risk on customers' attention (effect size = 0.088, SE = 0.026, 95% CI [0.050, 0.149]), enthusiasm (effect size = 0.078, SE = 0.020, 95% CI [0.045, 0.126]), and interaction (effect size = 0.098, SE = 0.023, 95% CI [0.060, 0.152]). Similarly, for low levels of health consciousness, the mediating effect of social distancing on customers' attention (effect size = 0.077, SE = 0.027, 95% CI [0.035, 0.139]), enthusiasm (effect size = 0.068, SE = 0.022, 95% CI [0.033, 0.119]), and interaction (effect size = 0.087, SE = 0.026, 95% CI [0.045, 0.143]) were still significant. There was no significant difference between high and low levels of health consciousness (attention: from 0.088 to 0.077; enthusiasm: from 0.078 to 0.068; interaction: from 0.098 to 0.087, Figure 3B). These results showed that the health consciousness did not moderate the mediating effects of social distancing on the relationship between perceived risk and customer–robot engagement. Thus, H4a, H4b, and H4c were not supported.
