Table 2.
Results of regression analysis.
| All scenarios | Scenarios involving an AI | Scenarios involving a human recruiter | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Outcomefairness | Processfairness | Inter-personaltreatment | Recom-mendationintention | Outcomefairness | Processfairness | Inter-personaltreatment | Recom-mendationintention | Outcomefairness | Processfairness | Inter-personaltreatment | Recom-mendationintention | |
| (Model 1) | (Model 2) | (Model 3) | (Model 4) | (Model 1a) | (Model 2a) | (Model 3a) | (Model 4a) | (Model 1b) | (Model 2b) | (Model 3b) | (Model 4b) | |
| AI with explanation | 0.323*** (0.060) | 0.394*** (0.091) | 0.526*** (0.087) | 0.372*** (0.100) | 0.336*** (0.065) | 0.404*** (0.101) | 0.546*** (0.092) | 0.383*** (0.105) | ||||
| Human without explanation | 0.323*** (0.061) | 0.416*** (0.092) | 0.372*** (0.087) | 0.262*** (0.100) | ||||||||
| Human with explanation | 0.371*** (0.061) | 0.614*** (0.092) | 0.561*** (0.087) | 0.437*** (0.100) | 0.051 (0.055) | 0.195** (0.081) | 0.185** (0.082) | 0.169* (0.095) | ||||
| Education (Ref. = compulsory school) | ||||||||||||
|
−0.160** (0.063) | −0.243** (0.095) | −0.203** (0.090) | −0.215** (0.104) | −0.048 (0.095) | −0.177 (0.148) | −0.068 (0.135) | −0.042 (0.154) | −0.284*** (0.082) | −0.313*** (0.120) | −0.334*** (0.121) | −0.387*** (0.140) |
|
−0.006 (0.079) | −0.157 (0.120) | −0.087 (0.114) | −0.199 (0.131) | 0.163 (0.119) | −0.034 (0.185) | 0.173 (0.168) | −0.032 (0.192) | −0.182* (0.104) | −0.287* (0.154) | −0.378** (0.154) | −0.384** (0.179) |
|
−0.199*** (0.073) | −0.398*** (0.110) | −0.335*** (0.104) | −0.369*** (0.120) | −0.185* (0.112) | −0.315* (0.173) | −0.204 (0.158) | −0.208 (0.180) | −0.219** (0.093) | −0.466*** (0.137) | −0.450*** (0.137) | −0.525*** (0.160) |
|
−0.179** (0.071) | −0.343*** (0.107) | −0.369*** (0.102) | −0.399*** (0.117) | −0.097 (0.110) | −0.300* (0.171) | −0.234 (0.156) | −0.240 (0.178) | −0.258*** (0.090) | −0.385*** (0.132) | −0.501*** (0.132) | −0.559*** (0.154) |
| Age | −0.005*** (0.002) | −0.009*** (0.003) | −0.004 (0.003) | −0.006** (0.003) | −0.007** (0.003) | −0.015*** (0.004) | −0.009** (0.004) | −0.010** (0.005) | −0.003 (0.002) | −0.004 (0.004) | 0.001 (0.004) | −0.003 (0.004) |
| Gender (woman) | −0.086** (0.043) | −0.228*** (0.066) | −0.192*** (0.062) | −0.222*** (0.072) | −0.133** (0.067) | −0.235** (0.104) | −0.219** (0.094) | −0.273** (0.108) | −0.038 (0.056) | −0.224*** (0.082) | −0.158* (0.082) | −0.165* (0.095) |
| Experienced discrimination | 0.030** (0.014) | 0.037* (0.021) | 0.048** (0.020) | 0.090*** (0.023) | 0.049** (0.022) | 0.042 (0.035) | 0.031 (0.032) | 0.075** (0.036) | 0.016 (0.018) | 0.039 (0.026) | 0.072*** (0.026) | 0.105*** (0.031) |
| Constant | 2.184*** (0.101) | 2.372*** (0.154) | 2.076*** (0.146) | 1.962*** (0.167) | 2.170*** (0.149) | 2.515*** (0.231) | 2.158*** (0.210) | 2.010*** (0.240) | 2.507*** (0.126) | 2.613*** (0.185) | 2.313*** (0.186) | 2.159*** (0.216) |
| Observations | 921 | 921 | 921 | 921 | 462 | 462 | 462 | 462 | 459 | 459 | 459 | 459 |
| R 2 | 0.082 | 0.094 | 0.092 | 0.076 | 0.095 | 0.079 | 0.099 | 0.067 | 0.046 | 0.072 | 0.080 | 0.087 |
| Adjusted R2 | 0.072 | 0.084 | 0.083 | 0.066 | 0.079 | 0.063 | 0.084 | 0.051 | 0.029 | 0.056 | 0.063 | 0.070 |
| Residual Std. error | 0.645 (df = 910) | 0.978 (df = 910) | 0.929 (df = 910) | 1.066 (df = 910) | 0.693 (df = 453) | 1.077 (df = 453) | 0.981 (df = 453) | 1.119 (df = 453) | 0.591 (df = 450) | 0.871 (df = 450) | 0.871 (df = 450) | 1.013 (df = 450) |
| F statistic | 8.138*** (df = 10; 910) | 9.431*** (df = 10; 910) | 9.275*** (df = 10; 910) | 7.521*** (df = 10; 910) | 5.952*** (df = 8; 453) | 4.870*** (df = 8; 453) | 6.251*** (df = 8; 453) | 4.093*** (df = 8; 453) | 2.722*** (df = 8; 450) | 4.389*** (df = 8; 450) | 4.875*** (df = 8; 450) | 5.329*** (df = 8; 450) |
*p, **p, ***p < 0.01. The table displays the unstandardized coefficients of OLS regression models with dummy variables for the different scenarios as independent variables (reference category is AI without explanation). Standard errors are given in parentheses. Models 1 to 4 are based on the pooled sample of all four scenarios, while Models 1a to 4a are based on the two pooled AI scenarios or Models 1b to 4b on the two pooled human scenarios.