For all means, higher numbers = more trust in humans, lower numbers = more trust in algorithms. T-tests are against the scale midpoint (trust both equally).
For each task I also check whether trust in algorithms vs. humans correlates with conservatism, reliance on feelings vs. facts, coastal vs. landlocked state, desire for control, and level knowledge of AI.
Personality analysis
setwd("~/Documents/Dropbox/Research/Adrian")
a<-read.csv ("AI_humans_performance_data.csv", header=T, sep=",")
a$cont<-(a$cont1_1+a$cont2_1)/2
a$knowAI<-(a$know + a$often)
a$tif<-a$tif_1+(101-a$tif_2)
coast<-subset(a, coast==1)
notcoast<-subset(a, coast==2)
t.test(a$personality_1, mu=50)
##
## One Sample t-test
##
## data: a$personality_1
## t = 7.5002, df = 407, p-value = 4.036e-13
## alternative hypothesis: true mean is not equal to 50
## 95 percent confidence interval:
## 57.28493 62.46017
## sample estimates:
## mean of x
## 59.87255
#liberals trust algorithms more
cor.test(a$personality_1, a$poli_1)
##
## Pearson's product-moment correlation
##
## data: a$personality_1 and a$poli_1
## t = -1.844, df = 406, p-value = 0.06591
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.18656936 0.00600228
## sample estimates:
## cor
## -0.09113546
#trusting feelings more makes you trust humans more
cor.test(a$personality_1, a$tif)
##
## Pearson's product-moment correlation
##
## data: a$personality_1 and a$tif
## t = 4.8937, df = 406, p-value = 1.43e-06
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.1421846 0.3256347
## sample estimates:
## cor
## 0.2360114
#desire for control - no correlation
cor.test(a$personality_1, a$cont)
##
## Pearson's product-moment correlation
##
## data: a$personality_1 and a$cont
## t = -0.30417, df = 406, p-value = 0.7612
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.11201460 0.08211085
## sample estimates:
## cor
## -0.01509411
#knowledge of AI
cor.test(a$personality_1, a$knowAI)
##
## Pearson's product-moment correlation
##
## data: a$personality_1 and a$knowAI
## t = -1.1973, df = 406, p-value = 0.2319
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.15550685 0.03798564
## sample estimates:
## cor
## -0.05931775
#coastal people trust algorithms more
t.test(coast$personality_1, notcoast$personality_1)
##
## Welch Two Sample t-test
##
## data: coast$personality_1 and notcoast$personality_1
## t = -2.1894, df = 379.01, p-value = 0.02918
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -10.9299699 -0.5868535
## sample estimates:
## mean of x mean of y
## 57.34043 63.09884
Driverless car
t.test(a$car_1, mu=50)
##
## One Sample t-test
##
## data: a$car_1
## t = -1.8383, df = 407, p-value = 0.06675
## alternative hypothesis: true mean is not equal to 50
## 95 percent confidence interval:
## 44.62879 50.18003
## sample estimates:
## mean of x
## 47.40441
#liberals trust algorithms more
cor.test(a$car_1, a$poli_1)
##
## Pearson's product-moment correlation
##
## data: a$car_1 and a$poli_1
## t = -3.3754, df = 406, p-value = 0.0008081
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.25815981 -0.06924204
## sample estimates:
## cor
## -0.165216
#trusting feelings more makes you trust humans more
cor.test(a$car_1, a$tif)
##
## Pearson's product-moment correlation
##
## data: a$car_1 and a$tif
## t = 5.3726, df = 406, p-value = 1.31e-07
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.1646700 0.3460646
## sample estimates:
## cor
## 0.2576359
#more desire for control, less trust of algorithms
cor.test(a$car_1, a$cont)
##
## Pearson's product-moment correlation
##
## data: a$car_1 and a$cont
## t = 4.249, df = 406, p-value = 2.665e-05
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.1114867 0.2974638
## sample estimates:
## cor
## 0.206338
#more knowledge of AI, more trust in algorithms
cor.test(a$car_1, a$knowAI)
##
## Pearson's product-moment correlation
##
## data: a$car_1 and a$knowAI
## t = -2.9123, df = 406, p-value = 0.003785
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.2368456 -0.0466130
## sample estimates:
## cor
## -0.1430503
#no difference for coastal people
t.test(coast$car_1, notcoast$car_1)
##
## Welch Two Sample t-test
##
## data: coast$car_1 and notcoast$car_1
## t = -0.76642, df = 372.41, p-value = 0.4439
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -7.779139 3.415755
## sample estimates:
## mean of x mean of y
## 46.35319 48.53488
Predicting joke funniness
t.test(a$joke_1, mu=50)
##
## One Sample t-test
##
## data: a$joke_1
## t = 13.908, df = 407, p-value < 2.2e-16
## alternative hypothesis: true mean is not equal to 50
## 95 percent confidence interval:
## 63.36390 67.76355
## sample estimates:
## mean of x
## 65.56373
#liberals - no diff.
cor.test(a$joke_1, a$poli_1)
##
## Pearson's product-moment correlation
##
## data: a$joke_1 and a$poli_1
## t = 0.3881, df = 406, p-value = 0.6981
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.07797298 0.11612496
## sample estimates:
## cor
## 0.01925743
#trusting feelings more makes you trust humans more
cor.test(a$joke_1, a$tif)
##
## Pearson's product-moment correlation
##
## data: a$joke_1 and a$tif
## t = 4.2952, df = 406, p-value = 2.186e-05
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.1137010 0.2995067
## sample estimates:
## cor
## 0.2084843
#desire for control
cor.test(a$joke_1, a$cont)
##
## Pearson's product-moment correlation
##
## data: a$joke_1 and a$cont
## t = -0.18096, df = 406, p-value = 0.8565
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.10597278 0.08818098
## sample estimates:
## cor
## -0.008980536
#knowledge of AI
cor.test(a$joke_1, a$knowAI)
##
## Pearson's product-moment correlation
##
## data: a$joke_1 and a$knowAI
## t = 0.075171, df = 406, p-value = 0.9401
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.09338781 0.10077878
## sample estimates:
## cor
## 0.003730644
#coastal
t.test(coast$joke_1, notcoast$joke_1)
##
## Welch Two Sample t-test
##
## data: coast$joke_1 and notcoast$joke_1
## t = -1.3366, df = 378.72, p-value = 0.1822
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -7.423925 1.415266
## sample estimates:
## mean of x mean of y
## 64.23404 67.23837
Cancer treatment plan
t.test(a$cancer_1, mu=50)
##
## One Sample t-test
##
## data: a$cancer_1
## t = -6.1953, df = 407, p-value = 1.425e-09
## alternative hypothesis: true mean is not equal to 50
## 95 percent confidence interval:
## 39.83932 44.73421
## sample estimates:
## mean of x
## 42.28676
#liberals trust algorithms more
cor.test(a$cancer_1, a$poli_1)
##
## Pearson's product-moment correlation
##
## data: a$cancer_1 and a$poli_1
## t = -1.899, df = 406, p-value = 0.05828
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.189189867 0.003285946
## sample estimates:
## cor
## -0.09382862
#trusting feelings more makes you trust humans more
cor.test(a$cancer_1, a$tif)
##
## Pearson's product-moment correlation
##
## data: a$cancer_1 and a$tif
## t = 3.1228, df = 406, p-value = 0.00192
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.05691622 0.24657293
## sample estimates:
## cor
## 0.1531546
#desire for control - no correlation
cor.test(a$cancer_1, a$cont)
##
## Pearson's product-moment correlation
##
## data: a$cancer_1 and a$cont
## t = 1.1487, df = 406, p-value = 0.2514
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.04039094 0.15315522
## sample estimates:
## cor
## 0.05691688
#knowledge of AI
cor.test(a$cancer_1, a$knowAI)
##
## Pearson's product-moment correlation
##
## data: a$cancer_1 and a$knowAI
## t = -2.3821, df = 406, p-value = 0.01767
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.21207296 -0.02055525
## sample estimates:
## cor
## -0.1174056
#coastal
t.test(coast$cancer_1, notcoast$cancer_1)
##
## Welch Two Sample t-test
##
## data: coast$cancer_1 and notcoast$cancer_1
## t = -0.28193, df = 348.8, p-value = 0.7782
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -5.751581 4.309374
## sample estimates:
## mean of x mean of y
## 41.88936 42.61047
Recommending movie
t.test(a$movie_1, mu=50)
##
## One Sample t-test
##
## data: a$movie_1
## t = 0.64972, df = 407, p-value = 0.5162
## alternative hypothesis: true mean is not equal to 50
## 95 percent confidence interval:
## 48.40134 53.17709
## sample estimates:
## mean of x
## 50.78922
#liberals trust algorithms more
cor.test(a$movie_1, a$poli_1)
##
## Pearson's product-moment correlation
##
## data: a$movie_1 and a$poli_1
## t = -3.0772, df = 406, p-value = 0.002231
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.24447208 -0.05468758
## sample estimates:
## cor
## -0.1509707
#trusting feelings more makes you trust humans more
cor.test(a$movie_1, a$tif)
##
## Pearson's product-moment correlation
##
## data: a$movie_1 and a$tif
## t = 3.121, df = 406, p-value = 0.001931
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.05682603 0.24648794
## sample estimates:
## cor
## 0.1530662
#desire for control
cor.test(a$movie_1, a$cont)
##
## Pearson's product-moment correlation
##
## data: a$movie_1 and a$cont
## t = 2.0802, df = 406, p-value = 0.03814
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.00566362 0.19780436
## sample estimates:
## cor
## 0.1026918
#knowledge of AI
cor.test(a$movie_1, a$knowAI)
##
## Pearson's product-moment correlation
##
## data: a$movie_1 and a$knowAI
## t = -1.6519, df = 406, p-value = 0.09933
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.17738630 0.01549877
## sample estimates:
## cor
## -0.08170881
#coastal
t.test(coast$movie_1, notcoast$movie_1)
##
## Welch Two Sample t-test
##
## data: coast$movie_1 and notcoast$movie_1
## t = 0.76401, df = 378.32, p-value = 0.4453
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -2.937666 6.671362
## sample estimates:
## mean of x mean of y
## 51.50638 49.63953
Diagnosisng psychological disorder
t.test(a$psych_1, mu=50)
##
## One Sample t-test
##
## data: a$psych_1
## t = 1.6061, df = 407, p-value = 0.109
## alternative hypothesis: true mean is not equal to 50
## 95 percent confidence interval:
## 49.57569 54.21352
## sample estimates:
## mean of x
## 51.89461
#liberals
cor.test(a$psych_1, a$poli_1)
##
## Pearson's product-moment correlation
##
## data: a$psych_1 and a$poli_1
## t = -0.20006, df = 406, p-value = 0.8415
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.10691007 0.08724024
## sample estimates:
## cor
## -0.009928488
#trusting feelings more makes you trust humans more
cor.test(a$psych_1, a$tif)
##
## Pearson's product-moment correlation
##
## data: a$psych_1 and a$tif
## t = 3.6817, df = 406, p-value = 0.000263
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.0841260 0.2720794
## sample estimates:
## cor
## 0.1797426
#desire for control - no correlation
cor.test(a$psych_1, a$cont)
##
## Pearson's product-moment correlation
##
## data: a$psych_1 and a$cont
## t = 0.80249, df = 406, p-value = 0.4227
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.05751165 0.13635301
## sample estimates:
## cor
## 0.03979518
#knowledge of AI
cor.test(a$psych_1, a$knowAI)
##
## Pearson's product-moment correlation
##
## data: a$psych_1 and a$knowAI
## t = -1.6294, df = 406, p-value = 0.104
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.1763088 0.0166108
## sample estimates:
## cor
## -0.08060382
#coastal
t.test(coast$psych_1, notcoast$psych_1)
##
## Welch Two Sample t-test
##
## data: coast$psych_1 and notcoast$psych_1
## t = 0.37658, df = 362.56, p-value = 0.7067
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -3.811587 5.617129
## sample estimates:
## mean of x mean of y
## 52.17021 51.26744
Deciding parole
t.test(a$parole_1, mu=50)
##
## One Sample t-test
##
## data: a$parole_1
## t = -4.3258, df = 407, p-value = 1.914e-05
## alternative hypothesis: true mean is not equal to 50
## 95 percent confidence interval:
## 41.61203 46.85366
## sample estimates:
## mean of x
## 44.23284
#liberals
cor.test(a$parole_1, a$poli_1)
##
## Pearson's product-moment correlation
##
## data: a$parole_1 and a$poli_1
## t = 0.18628, df = 406, p-value = 0.8523
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.08791918 0.10623366
## sample estimates:
## cor
## 0.009244366
#trusting feelings more makes you trust humans more
cor.test(a$parole_1, a$tif)
##
## Pearson's product-moment correlation
##
## data: a$parole_1 and a$tif
## t = 3.0194, df = 406, p-value = 0.002692
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.05185852 0.24180270
## sample estimates:
## cor
## 0.148197
#desire for control
cor.test(a$parole_1, a$cont)
##
## Pearson's product-moment correlation
##
## data: a$parole_1 and a$cont
## t = 2.6381, df = 406, p-value = 0.008659
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.03315022 0.22407759
## sample estimates:
## cor
## 0.1298171
#knowledge of AI
cor.test(a$parole_1, a$knowAI)
##
## Pearson's product-moment correlation
##
## data: a$parole_1 and a$knowAI
## t = -0.69424, df = 406, p-value = 0.4879
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.13108042 0.06286079
## sample estimates:
## cor
## -0.03443399
#coastal people trust algorithms more
t.test(coast$parole_1, notcoast$parole_1)
##
## Welch Two Sample t-test
##
## data: coast$parole_1 and notcoast$parole_1
## t = -0.15336, df = 362.57, p-value = 0.8782
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -5.747213 4.915645
## sample estimates:
## mean of x mean of y
## 43.94468 44.36047
Admitting students to college
t.test(a$gpa_1, mu=50)
##
## One Sample t-test
##
## data: a$gpa_1
## t = -7.7393, df = 407, p-value = 7.979e-14
## alternative hypothesis: true mean is not equal to 50
## 95 percent confidence interval:
## 36.65161 42.05917
## sample estimates:
## mean of x
## 39.35539
#liberals
cor.test(a$gpa_1, a$poli_1)
##
## Pearson's product-moment correlation
##
## data: a$gpa_1 and a$poli_1
## t = 0.49121, df = 406, p-value = 0.6235
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.0728862 0.1211688
## sample estimates:
## cor
## 0.02437088
#trusting feelings more makes you trust humans more
cor.test(a$gpa_1, a$tif)
##
## Pearson's product-moment correlation
##
## data: a$gpa_1 and a$tif
## t = 4.3415, df = 406, p-value = 1.789e-05
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.1159184 0.3015507
## sample estimates:
## cor
## 0.2106326
#desire for control
cor.test(a$gpa_1, a$cont)
##
## Pearson's product-moment correlation
##
## data: a$gpa_1 and a$cont
## t = 2.3776, df = 406, p-value = 0.01789
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.02033304 0.21186065
## sample estimates:
## cor
## 0.1171864
#knowledge of AI
cor.test(a$gpa_1, a$knowAI)
##
## Pearson's product-moment correlation
##
## data: a$gpa_1 and a$knowAI
## t = -0.52352, df = 406, p-value = 0.6009
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.1227481 0.0712914
## sample estimates:
## cor
## -0.02597301
#coastal people trust algorithms more
t.test(coast$gpa_1, notcoast$gpa_1)
##
## Welch Two Sample t-test
##
## data: coast$gpa_1 and notcoast$gpa_1
## t = 1.1582, df = 368.06, p-value = 0.2475
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -2.246792 8.685584
## sample estimates:
## mean of x mean of y
## 40.59149 37.37209
Hiring employees
t.test(a$job_1, mu=50)
##
## One Sample t-test
##
## data: a$job_1
## t = -6.0623, df = 407, p-value = 3.064e-09
## alternative hypothesis: true mean is not equal to 50
## 95 percent confidence interval:
## 38.30879 44.03435
## sample estimates:
## mean of x
## 41.17157
#liberals
cor.test(a$job_1, a$poli_1)
##
## Pearson's product-moment correlation
##
## data: a$job_1 and a$poli_1
## t = -0.49959, df = 406, p-value = 0.6176
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.1215789 0.0724722
## sample estimates:
## cor
## -0.02478683
#trusting feelings more makes you trust humans more
cor.test(a$job_1, a$tif)
##
## Pearson's product-moment correlation
##
## data: a$job_1 and a$tif
## t = 5.3101, df = 406, p-value = 1.809e-07
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## 0.1617513 0.3434223
## sample estimates:
## cor
## 0.2548341
#desire for control
cor.test(a$job_1, a$cont)
##
## Pearson's product-moment correlation
##
## data: a$job_1 and a$cont
## t = 1.9151, df = 406, p-value = 0.05619
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.002491146 0.189956110
## sample estimates:
## cor
## 0.09461637
#knowledge of AI
cor.test(a$job_1, a$knowAI)
##
## Pearson's product-moment correlation
##
## data: a$job_1 and a$knowAI
## t = -1.1752, df = 406, p-value = 0.2406
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
## -0.15443783 0.03907934
## sample estimates:
## cor
## -0.05822621
#coastal people trust algorithms more
t.test(coast$job_1, notcoast$job_1)
##
## Welch Two Sample t-test
##
## data: coast$job_1 and notcoast$job_1
## t = 0.39353, df = 359.92, p-value = 0.6942
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
## -4.672991 7.011091
## sample estimates:
## mean of x mean of y
## 41.57021 40.40116