There is a total of 353 entries in the dataset. 18 participants failed the attention check or the comprehension check. The valid number of entries is n = 335.
Unlike S3-mediation.Rmd, which discovers candidate mediators empirically, this file tests the specific mediation hypotheses named in the preregistration (section 2) against the comfort DV (favorability) and reports a pass/fail-style verdict for each.
| ID | Preregistered hypothesis | Comparison | Mediator(s) |
|---|---|---|---|
| H4a | AI-trained doctors viewed less favorably than non-AI-trained doctors, because AI-trained doctors are perceived lower in ability and autonomy | no_ai vs. yes_ai (reference) |
ability, autonomy |
| H4b | AI-trained doctors viewed less favorably than doctors with unknown AI history, because AI-trained doctors are perceived lower in ability and autonomy | control vs. yes_ai (reference) |
ability, autonomy |
| H4c | Non-AI-trained doctors viewed more favorably than doctors with unknown AI history, because non-AI-trained doctors are perceived higher in ability and autonomy | no_ai vs. control (reference) |
ability, autonomy |
| H5a | Peer-tutor-trained doctors viewed less favorably than non-peer-tutor-trained doctors, because peer-trained doctors are perceived lower in ability | no_human vs. yes_human (reference) |
ability |
| H5b | Peer-tutor-trained doctors viewed less favorably than doctors with unknown tutor history, because peer-trained doctors are perceived lower in ability | control vs. yes_human (reference) |
ability |
Note: the preregistration does not name a mediation hypothesis
analogous to H4c for the human-training arm (it explicitly does
not expect no_human to outperform
control), so there is no “H5c” tested here.
Reference-group sign convention: for every comparison above, the reference (lower-favorability) group is coded 0 and the higher-favorability group is coded 1, so that a positive effect always means “in the direction the preregistration predicts.” A hypothesis is called:
Each named mediator here is tested as part of one parallel mediation model per comparison (both ability and autonomy together for the AI hypotheses, matching the “and” in the hypothesis wording), so the indirect effects reported are each mediator’s unique contribution controlling for the other.
Importing Hayes’ PROCESS macro for R (v4.1.1) so
process() is available below.
##
## ********************* PROCESS for R Version 4.1.1 *********************
##
## Written by Andrew F. Hayes, Ph.D. www.afhayes.com
## Documentation available in Hayes (2022). www.guilford.com/p/hayes3
##
## ***********************************************************************
##
## PROCESS is now ready for use.
## Copyright 2022 by Andrew F. Hayes ALL RIGHTS RESERVED
## Workshop schedule at http://haskayne.ucalgary.ca/CCRAM
## yes_ai reference) – H4a, H4bParallel mediation model with both preregistered mediators (ability,
autonomy), Y = comfort. Reference: yes_ai (X1 =
no_ai vs. yes_ai, X2 = control
vs. yes_ai).
##
## ********************* PROCESS for R Version 4.1.1 *********************
##
## Written by Andrew F. Hayes, Ph.D. www.afhayes.com
## Documentation available in Hayes (2022). www.guilford.com/p/hayes3
##
## ***********************************************************************
##
## Model : 4
## Y : comfort
## X : ai_training_1
## M1 : ability
## M2 : autonomy
##
## Sample size: 196
##
## Custom seed: 1234
##
## Coding of categorical X variable for analysis:
## ai_training_1 X1 X2
## 1.0000 0.0000 0.0000
## 2.0000 1.0000 0.0000
## 3.0000 0.0000 1.0000
##
## ***********************************************************************
## Outcome Variable: ability
##
## Model Summary:
## R R-sq MSE F df1 df2 p
## 0.5136 0.2638 0.9083 34.5844 2.0000 193.0000 0.0000
##
## Model:
## coeff se t p LLCI ULCI
## constant 4.1364 0.1173 35.2594 0.0000 3.9050 4.3677
## X1 1.3564 0.1641 8.2660 0.0000 1.0327 1.6800
## X2 0.8281 0.1693 4.8923 0.0000 0.4943 1.1620
##
## ***********************************************************************
## Outcome Variable: autonomy
##
## Model Summary:
## R R-sq MSE F df1 df2 p
## 0.5322 0.2833 1.1100 38.1389 2.0000 193.0000 0.0000
##
## Model:
## coeff se t p LLCI ULCI
## constant 3.5177 0.1297 27.1253 0.0000 3.2619 3.7735
## X1 1.5838 0.1814 8.7311 0.0000 1.2260 1.9415
## X2 0.7747 0.1871 4.1400 0.0001 0.4056 1.1437
##
## ***********************************************************************
## Outcome Variable: comfort
##
## Model Summary:
## R R-sq MSE F df1 df2 p
## 0.8769 0.7689 0.4257 158.8770 4.0000 191.0000 0.0000
##
## Model:
## coeff se t p LLCI ULCI
## constant -0.3161 0.2216 -1.4268 0.1553 -0.7531 0.1209
## X1 0.1488 0.1356 1.0975 0.2738 -0.1186 0.4162
## X2 -0.0148 0.1233 -0.1202 0.9044 -0.2581 0.2284
## ability 0.9544 0.0645 14.7936 0.0000 0.8272 1.0817
## autonomy 0.0961 0.0584 1.6461 0.1014 -0.0190 0.2112
##
## ************************ TOTAL EFFECT MODEL ***************************
## Outcome Variable: comfort
##
## Model Summary:
## R R-sq MSE F df1 df2 p
## 0.4942 0.2442 1.3778 31.1821 2.0000 193.0000 0.0000
##
## Model:
## coeff se t p LLCI ULCI
## constant 3.9697 0.1445 27.4751 0.0000 3.6847 4.2547
## X1 1.5955 0.2021 7.8948 0.0000 1.1969 1.9941
## X2 0.8500 0.2085 4.0771 0.0001 0.4388 1.2612
##
## ***********************************************************************
## Bootstrapping in progress. Please wait.
##
## ************ TOTAL, DIRECT, AND INDIRECT EFFECTS OF X ON Y ************
##
## Relative total effects of X on Y:
## effect se t p LLCI ULCI
## X1 1.5955 0.2021 7.8948 0.0000 1.1969 1.9941
## X2 0.8500 0.2085 4.0771 0.0001 0.4388 1.2612
##
## Omnibus test of total effect of X on Y:
## R2-chng F df1 df2 p
## 0.2442 31.1821 2.0000 193.0000 0.0000
## ----------
##
## Relative direct effects of X on Y:
## effect se t p LLCI ULCI
## X1 0.1488 0.1356 1.0975 0.2738 -0.1186 0.4162
## X2 -0.0148 0.1233 -0.1202 0.9044 -0.2581 0.2284
##
## Omnibus test of direct effect of X on Y:
## R2-chng F df1 df2 p
## 0.0025 1.0180 2.0000 191.0000 0.3633
##
## ----------
##
## Relative indirect effects of X on Y:
##
## ai_training_1 -> ability -> comfort
##
## Effect BootSE BootLLCI BootULCI
## X1 1.2946 0.1919 0.9276 1.6846
## X2 0.7904 0.1903 0.4272 1.1780
##
## Normal theory test for relative indirect effects:
## Effect se Z p
## X1 1.2946 0.1797 7.2035 0.0000
## X2 0.7904 0.1705 4.6353 0.0000
##
## ai_training_1 -> autonomy -> comfort
##
## Effect BootSE BootLLCI BootULCI
## X1 0.1522 0.0889 -0.0145 0.3351
## X2 0.0744 0.0482 -0.0068 0.1811
##
## Normal theory test for relative indirect effects:
## Effect se Z p
## X1 0.1522 0.0947 1.6075 0.1079
## X2 0.0744 0.0499 1.4925 0.1356
##
## ******************** ANALYSIS NOTES AND ERRORS ************************
##
## Level of confidence for all confidence intervals in output: 95
##
## Number of bootstraps for percentile bootstrap confidence intervals: 10000
##
## NOTE: Some cases with missing data were deleted. The number of deleted cases was: 2##
## ==============================================================================
## H4a: AI-trained doctors viewed less favorably than non-AI-trained doctors, because AI-trained doctors are perceived lower in ability and autonomy.
## Comparison: no_ai vs. yes_ai (reference) | Y = comfort
## ==============================================================================
##
## Total effect: b = 1.5955, p = 0.0000 [predicted: positive & significant]
## -> base favorability prediction CONFIRMED
##
## Indirect effect via ability : b = 1.2946, 95% Boot CI [ 0.9276, 1.6846], normal-theory p = 0.0000 -> PREDICTED DIRECTION, SIGNIFICANT
## Indirect effect via autonomy : b = 0.1522, 95% Boot CI [ -0.0145, 0.3351], normal-theory p = 0.1079 -> NOT SIGNIFICANT
##
## VERDICT: PARTIALLY SUPPORTED##
## ==============================================================================
## H4b: AI-trained doctors viewed less favorably than doctors with unknown AI history, because AI-trained doctors are perceived lower in ability and autonomy.
## Comparison: control vs. yes_ai (reference) | Y = comfort
## ==============================================================================
##
## Total effect: b = 0.8500, p = 0.0001 [predicted: positive & significant]
## -> base favorability prediction CONFIRMED
##
## Indirect effect via ability : b = 0.7904, 95% Boot CI [ 0.4272, 1.1780], normal-theory p = 0.0000 -> PREDICTED DIRECTION, SIGNIFICANT
## Indirect effect via autonomy : b = 0.0744, 95% Boot CI [ -0.0068, 0.1811], normal-theory p = 0.1356 -> NOT SIGNIFICANT
##
## VERDICT: PARTIALLY SUPPORTEDcontrol reference) – H4cno_ai vs. control only (yes_ai
excluded from this comparison). Parallel mediation model with both
preregistered mediators (ability, autonomy), Y = comfort.
##
## ********************* PROCESS for R Version 4.1.1 *********************
##
## Written by Andrew F. Hayes, Ph.D. www.afhayes.com
## Documentation available in Hayes (2022). www.guilford.com/p/hayes3
##
## ***********************************************************************
##
## Model : 4
## Y : comfort
## X : x_no_ai_vs_control
## M1 : ability
## M2 : autonomy
##
## Sample size: 130
##
## Custom seed: 1234
##
##
## ***********************************************************************
## Outcome Variable: ability
##
## Model Summary:
## R R-sq MSE F df1 df2 p
## 0.3684 0.1357 0.4495 20.1031 1.0000 128.0000 0.0000
##
## Model:
## coeff se t p LLCI
## constant 4.9645 0.0858 57.8355 0.0000 4.7946
## x_no_ai_vs_control 0.5283 0.1178 4.4837 0.0000 0.2951
## ULCI
## constant 5.1343
## x_no_ai_vs_control 0.7614
##
## ***********************************************************************
## Outcome Variable: autonomy
##
## Model Summary:
## R R-sq MSE F df1 df2 p
## 0.4227 0.1787 0.7612 27.8435 1.0000 128.0000 0.0000
##
## Model:
## coeff se t p LLCI
## constant 4.2923 0.1117 38.4240 0.0000 4.0713
## x_no_ai_vs_control 0.8091 0.1533 5.2767 0.0000 0.5057
## ULCI
## constant 4.5134
## x_no_ai_vs_control 1.1125
##
## ***********************************************************************
## Outcome Variable: comfort
##
## Model Summary:
## R R-sq MSE F df1 df2 p
## 0.7695 0.5921 0.4078 60.9673 3.0000 126.0000 0.0000
##
## Model:
## coeff se t p LLCI
## constant -0.1037 0.4298 -0.2414 0.8097 -0.9542
## x_no_ai_vs_control 0.2218 0.1253 1.7708 0.0790 -0.0261
## ability 0.9921 0.1008 9.8469 0.0000 0.7927
## autonomy -0.0005 0.0774 -0.0060 0.9952 -0.1537
## ULCI
## constant 0.7468
## x_no_ai_vs_control 0.4697
## ability 1.1915
## autonomy 0.1527
##
## ************************ TOTAL EFFECT MODEL ***************************
## Outcome Variable: comfort
##
## Model Summary:
## R R-sq MSE F df1 df2 p
## 0.3780 0.1429 0.8435 21.3343 1.0000 128.0000 0.0000
##
## Model:
## coeff se t p LLCI
## constant 4.8197 0.1176 40.9855 0.0000 4.5870
## x_no_ai_vs_control 0.7455 0.1614 4.6189 0.0000 0.4262
## ULCI
## constant 5.0524
## x_no_ai_vs_control 1.0649
##
## ***********************************************************************
## Bootstrapping in progress. Please wait.
##
## ************ TOTAL, DIRECT, AND INDIRECT EFFECTS OF X ON Y ************
##
## Total effect of X on Y:
## effect se t p LLCI ULCI
## 0.7455 0.1614 4.6189 0.0000 0.4262 1.0649
##
## Direct effect of X on Y:
## effect se t p LLCI ULCI
## 0.2218 0.1253 1.7708 0.0790 -0.0261 0.4697
##
## Indirect effect(s) of X on Y:
## Effect BootSE BootLLCI BootULCI
## TOTAL 0.5237 0.1401 0.2618 0.8099
## ability 0.5241 0.1388 0.2712 0.8159
## autonomy -0.0004 0.0637 -0.1302 0.1248
##
## Normal theory test for indirect effect(s):
## Effect se Z p
## ability 0.5241 0.1290 4.0632 0.0000
## autonomy -0.0004 0.0638 -0.0059 0.9953
##
## ******************** ANALYSIS NOTES AND ERRORS ************************
##
## Level of confidence for all confidence intervals in output: 95
##
## Number of bootstraps for percentile bootstrap confidence intervals: 10000
##
## NOTE: Some cases with missing data were deleted. The number of deleted cases was: 1##
## ==============================================================================
## H4c: Non-AI-trained doctors viewed more favorably than doctors with unknown AI history, because non-AI-trained doctors are perceived higher in ability and autonomy.
## Comparison: no_ai vs. control (reference) | Y = comfort
## ==============================================================================
##
## Total effect: b = 0.7455, p = 0.0000 [predicted: positive & significant]
## -> base favorability prediction CONFIRMED
##
## Indirect effect via ability : b = 0.5241, 95% Boot CI [ 0.2712, 0.8159], normal-theory p = 0.0000 -> PREDICTED DIRECTION, SIGNIFICANT
## Indirect effect via autonomy : b = -0.0004, 95% Boot CI [ -0.1302, 0.1248], normal-theory p = 0.9953 -> NOT SIGNIFICANT
##
## VERDICT: PARTIALLY SUPPORTEDyes_human reference) – H5a, H5bSimple mediation model with the one preregistered mediator (ability),
Y = comfort. Reference: yes_human (X1 =
no_human vs. yes_human, X2 =
control vs. yes_human).
##
## ********************* PROCESS for R Version 4.1.1 *********************
##
## Written by Andrew F. Hayes, Ph.D. www.afhayes.com
## Documentation available in Hayes (2022). www.guilford.com/p/hayes3
##
## ***********************************************************************
##
## Model : 4
## Y : comfort
## X : human_training_1
## M : ability
##
## Sample size: 198
##
## Custom seed: 1234
##
## Coding of categorical X variable for analysis:
## human_training_1 X1 X2
## 1.0000 0.0000 0.0000
## 2.0000 1.0000 0.0000
## 3.0000 0.0000 1.0000
##
## ***********************************************************************
## Outcome Variable: ability
##
## Model Summary:
## R R-sq MSE F df1 df2 p
## 0.1422 0.0202 0.9246 2.0135 2.0000 195.0000 0.1363
##
## Model:
## coeff se t p LLCI ULCI
## constant 4.8920 0.1141 42.8680 0.0000 4.6670 5.1171
## X1 -0.2506 0.1644 -1.5242 0.1291 -0.5749 0.0737
## X2 0.0725 0.1679 0.4317 0.6665 -0.2586 0.4035
##
## ***********************************************************************
## Outcome Variable: comfort
##
## Model Summary:
## R R-sq MSE F df1 df2 p
## 0.8454 0.7147 0.4733 161.9922 3.0000 194.0000 0.0000
##
## Model:
## coeff se t p LLCI ULCI
## constant -0.5760 0.2636 -2.1851 0.0301 -1.0959 -0.0561
## X1 -0.2394 0.1183 -2.0232 0.0444 -0.4728 -0.0060
## X2 -0.0628 0.1202 -0.5228 0.6017 -0.2998 0.1742
## ability 1.0995 0.0512 21.4605 0.0000 0.9985 1.2006
##
## ************************ TOTAL EFFECT MODEL ***************************
## Outcome Variable: comfort
##
## Model Summary:
## R R-sq MSE F df1 df2 p
## 0.1934 0.0374 1.5886 3.7869 2.0000 195.0000 0.0244
##
## Model:
## coeff se t p LLCI ULCI
## constant 4.8028 0.1496 32.1079 0.0000 4.5078 5.0978
## X1 -0.5149 0.2155 -2.3894 0.0178 -0.9400 -0.0899
## X2 0.0169 0.2200 0.0766 0.9390 -0.4171 0.4508
##
## ***********************************************************************
## Bootstrapping in progress. Please wait.
##
## ************ TOTAL, DIRECT, AND INDIRECT EFFECTS OF X ON Y ************
##
## Relative total effects of X on Y:
## effect se t p LLCI ULCI
## X1 -0.5149 0.2155 -2.3894 0.0178 -0.9400 -0.0899
## X2 0.0169 0.2200 0.0766 0.9390 -0.4171 0.4508
##
## Omnibus test of total effect of X on Y:
## R2-chng F df1 df2 p
## 0.0374 3.7869 2.0000 195.0000 0.0244
## ----------
##
## Relative direct effects of X on Y:
## effect se t p LLCI ULCI
## X1 -0.2394 0.1183 -2.0232 0.0444 -0.4728 -0.0060
## X2 -0.0628 0.1202 -0.5228 0.6017 -0.2998 0.1742
##
## Omnibus test of direct effect of X on Y:
## R2-chng F df1 df2 p
## 0.0063 2.1586 2.0000 194.0000 0.1183
##
## ----------
##
## Relative indirect effects of X on Y:
##
## human_training_1 -> ability -> comfort
##
## Effect BootSE BootLLCI BootULCI
## X1 -0.2755 0.1927 -0.6562 0.1035
## X2 0.0797 0.1725 -0.2560 0.4210
##
## Normal theory test for relative indirect effects:
## Effect se Z p
## X1 -0.2755 0.1814 -1.5187 0.1288
## X2 0.0797 0.1848 0.4311 0.6664
##
## ******************** ANALYSIS NOTES AND ERRORS ************************
##
## Level of confidence for all confidence intervals in output: 95
##
## Number of bootstraps for percentile bootstrap confidence intervals: 10000
##
## NOTE: Some cases with missing data were deleted. The number of deleted cases was: 1##
## ==============================================================================
## H5a: Peer-tutor-trained doctors viewed less favorably than non-peer-tutor-trained doctors, because peer-trained doctors are perceived lower in ability.
## Comparison: no_human vs. yes_human (reference) | Y = comfort
## ==============================================================================
##
## Total effect: b = -0.5149, p = 0.0178 [predicted: positive & significant]
## -> base favorability prediction SIGNIFICANT BUT IN THE OPPOSITE DIRECTION
##
## Indirect effect via ability : b = -0.2755, 95% Boot CI [ -0.6562, 0.1035], normal-theory p = 0.1288 -> NOT SIGNIFICANT
##
## VERDICT: NOT SUPPORTED (base favorability effect SIGNIFICANT BUT REVERSED)##
## ==============================================================================
## H5b: Peer-tutor-trained doctors viewed less favorably than doctors with unknown tutor history, because peer-trained doctors are perceived lower in ability.
## Comparison: control vs. yes_human (reference) | Y = comfort
## ==============================================================================
##
## Total effect: b = 0.0169, p = 0.9390 [predicted: positive & significant]
## -> base favorability prediction NOT confirmed (not significant)
##
## Indirect effect via ability : b = 0.0797, 95% Boot CI [ -0.2560, 0.4210], normal-theory p = 0.6664 -> NOT SIGNIFICANT
##
## VERDICT: NOT SUPPORTED (base favorability effect not confirmed)| Hypothesis | Comparison | Total effect (b, p) | Mediators | Mediators confirmed | Verdict |
|---|---|---|---|---|---|
| H4a | no_ai vs. yes_ai (reference) | 1.595, p = 0.0000 | ability, autonomy | ability | PARTIALLY SUPPORTED |
| H4b | control vs. yes_ai (reference) | 0.850, p = 0.0001 | ability, autonomy | ability | PARTIALLY SUPPORTED |
| H4c | no_ai vs. control (reference) | 0.746, p = 0.0000 | ability, autonomy | ability | PARTIALLY SUPPORTED |
| H5a | no_human vs. yes_human (reference) | -0.515, p = 0.0178 | ability | none | NOT SUPPORTED (base favorability effect SIGNIFICANT BUT REVERSED) |
| H5b | control vs. yes_human (reference) | 0.017, p = 0.9390 | ability | none | NOT SUPPORTED (base favorability effect not confirmed) |