Looking at the reliance on feelings manipulation first

No difference in trust for any of the tasks between conditions

t.test(hiT$personality_1, loT$personality_1)
## 
##  Welch Two Sample t-test
## 
## data:  hiT$personality_1 and loT$personality_1
## t = 1.1455, df = 380.46, p-value = 0.2527
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -2.242649  8.502319
## sample estimates:
## mean of x mean of y 
##  61.95146  58.82162
t.test(hiT$joke_1, loT$joke_1)
## 
##  Welch Two Sample t-test
## 
## data:  hiT$joke_1 and loT$joke_1
## t = 0.3831, df = 388.66, p-value = 0.7019
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -3.699916  5.490733
## sample estimates:
## mean of x mean of y 
##  58.87379  57.97838
t.test(hiT$car_1, loT$car_1)
## 
##  Welch Two Sample t-test
## 
## data:  hiT$car_1 and loT$car_1
## t = 1.2887, df = 381.96, p-value = 0.1983
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -2.071818  9.954263
## sample estimates:
## mean of x mean of y 
##  49.18447  45.24324
t.test(hiT$movie_1, loT$movie_1)
## 
##  Welch Two Sample t-test
## 
## data:  hiT$movie_1 and loT$movie_1
## t = -0.19939, df = 382.79, p-value = 0.8421
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -5.423686  4.424945
## sample estimates:
## mean of x mean of y 
##  46.20874  46.70811
t.test(hiT$psych_1, loT$psych_1)
## 
##  Welch Two Sample t-test
## 
## data:  hiT$psych_1 and loT$psych_1
## t = -0.83495, df = 384.46, p-value = 0.4043
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -7.355502  2.970459
## sample estimates:
## mean of x mean of y 
##  51.79126  53.98378
t.test(hiT$trust, loT$trust)
## 
##  Welch Two Sample t-test
## 
## data:  hiT$trust and loT$trust
## t = 0.61364, df = 387.19, p-value = 0.5398
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -2.325052  4.434882
## sample estimates:
## mean of x mean of y 
##  53.60194  52.54703

the manipulation did succeed in changing reliance on feelings, but not trust in algorithms.. even though reliance on feelings and trust in algorithms are correlated. we must be missing something…

t.test(hiT$tif, loT$tif)
## 
##  Welch Two Sample t-test
## 
## data:  hiT$tif and loT$tif
## t = 4.2505, df = 378.28, p-value = 2.69e-05
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##   4.77004 12.98207
## sample estimates:
## mean of x mean of y 
##  58.44903  49.57297
cor.test(a$tif, a$trust)
## 
##  Pearson's product-moment correlation
## 
## data:  a$tif and a$trust
## t = 10.329, df = 1002, p-value < 2.2e-16
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.2531963 0.3650692
## sample estimates:
##       cor 
## 0.3102063

snowflake manipulation - no effect on trust nor on need for uniqueness

hiU<-subset(a, hiUNI!="NA")
loU<-subset(a, loUNI!="NA")

t.test(hiU$personality_1, loU$personality_1)
## 
##  Welch Two Sample t-test
## 
## data:  hiU$personality_1 and loU$personality_1
## t = 0.18005, df = 586.86, p-value = 0.8572
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -3.831489  4.604883
## sample estimates:
## mean of x mean of y 
##  59.74086  59.35417
t.test(hiU$joke_1, loU$joke_1)
## 
##  Welch Two Sample t-test
## 
## data:  hiU$joke_1 and loU$joke_1
## t = -1.1834, df = 584.01, p-value = 0.2371
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -5.816171  1.442624
## sample estimates:
## mean of x mean of y 
##  60.90698  63.09375
t.test(hiU$car_1, loU$car_1)
## 
##  Welch Two Sample t-test
## 
## data:  hiU$car_1 and loU$car_1
## t = 1.2448, df = 586.92, p-value = 0.2137
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -1.849894  8.253456
## sample estimates:
## mean of x mean of y 
##  52.29900  49.09722
t.test(hiU$movie_1, loU$movie_1)
## 
##  Welch Two Sample t-test
## 
## data:  hiU$movie_1 and loU$movie_1
## t = -0.64079, df = 585.05, p-value = 0.5219
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -5.259755  2.671946
## sample estimates:
## mean of x mean of y 
##  43.58804  44.88194
t.test(hiU$psych_1, loU$psych_1)
## 
##  Welch Two Sample t-test
## 
## data:  hiU$psych_1 and loU$psych_1
## t = 1.2459, df = 586.18, p-value = 0.2133
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -1.578435  7.055756
## sample estimates:
## mean of x mean of y 
##  59.94352  57.20486
t.test(hiU$trust, loU$trust)
## 
##  Welch Two Sample t-test
## 
## data:  hiU$trust and loU$trust
## t = 0.41242, df = 586.95, p-value = 0.6802
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -2.141790  3.280374
## sample estimates:
## mean of x mean of y 
##  55.29568  54.72639
t.test(hiU$nfu, loU$nfu)
## 
##  Welch Two Sample t-test
## 
## data:  hiU$nfu and loU$nfu
## t = -0.60527, df = 545.16, p-value = 0.5453
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -0.12903900  0.06824872
## sample estimates:
## mean of x mean of y 
##  5.437137  5.467532

this is odd - agreement with the arguments in the manipulation is correlated with more trust in algorithms in BOTH conditions

cor.test(a$hiUNI, a$trust)
## 
##  Pearson's product-moment correlation
## 
## data:  a$hiUNI and a$trust
## t = -4.4595, df = 299, p-value = 1.164e-05
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.3528185 -0.1406455
## sample estimates:
##        cor 
## -0.2497271
cor.test(a$loUNI, a$trust)
## 
##  Pearson's product-moment correlation
## 
## data:  a$loUNI and a$trust
## t = -2.3582, df = 286, p-value = 0.01904
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.24969861 -0.02289087
## sample estimates:
##        cor 
## -0.1381049

need for uniqueness isn’t correlated with trust in algorithms.

cor.test(a$nfu, a$trust)
## 
##  Pearson's product-moment correlation
## 
## data:  a$nfu and a$trust
## t = -0.27561, df = 926, p-value = 0.7829
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.07336813  0.05532967
## sample estimates:
##          cor 
## -0.009056735

higher need for cognition is correlated with more trust in algorithms (less trust in humans). the reliance on feelings manipulation also successfully manipulated need for cognition.

a$nfc<-(a$Q56+a$Q58+(6-a$Q60) + (6-a$Q62) + (6-a$Q64) + a$Q66 + (6-a$Q68) + (6-a$Q70) + (6-a$Q72) + a$Q74 + a$Q76 + a$Q80 + a$Q82 + a$Q84 + (6-a$Q86) + (6-a$Q88) + a$Q90)/16

cor.test(a$nfc, a$trust)
## 
##  Pearson's product-moment correlation
## 
## data:  a$nfc and a$trust
## t = -2.6949, df = 896, p-value = 0.007173
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.15418329 -0.02438922
## sample estimates:
##         cor 
## -0.08966695
hiT<-subset(a, hitif==1)
hiT<-subset(hiT, ch1==4 & ch2==4 & ch3==4)
loT<-subset(a, lotif==1)
loT<-subset(loT, ch4==1 & ch5==1 & ch6==2)

t.test(hiT$nfc, loT$nfc)
## 
##  Welch Two Sample t-test
## 
## data:  hiT$nfc and loT$nfc
## t = -2.3629, df = 344.38, p-value = 0.01869
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -0.39005283 -0.03567934
## sample estimates:
## mean of x mean of y 
##  3.649930  3.862796