Table 3 Proportion of judgments that correctly attributed the computer-generated passages to the computer.

From: Attributions toward artificial agents in a modified Moral Turing Test

Scenario

Percentage correctly choosing computer

Percentage choosing computer [95% CI]

z-score (p)

Tried on shirt in store (C)

82.17***

77.46–86.31

10.82 (< 0.001)

T-shirt at funeral (C)

81.12***

76.32–85.37

10.47 (< 0.001)

Robbed at gunpoint (M)

79.02***

74.06–83.47

9.76 (< 0.001)

Charged credit card (M)

77.27***

72.19–81.87

9.17 (< 0.001)

Hit pizza boy (M)

74.83***

69.59–79.62

8.34 (< 0.001)

Ate at restaurant with fingers (C)

72.03***

66.65–77.02

7.39 (< 0.001)

Man wore skirt to office (C)

65.73***

60.11–71.08

5.26 (< 0.001)

Ate in boss’ office (C)

61.54***

55.82–67.06

3.84 (< 0.001)

Shot dog (M)

58.74**

52.98–64.35

2.90 (0.004)

Set fire (M)

58.04

52.27–63.67

2.66 (0.008)

  1. Results represent the percentage of source attribution judgments that correctly identified the computer source of the passages. One-sample binomial tests were conducted with a null hypothesized median of 50% and used Bonferroni corrections (alpha = 0.005) for multiple tests. Scenarios involving moral transgressions are denoted by (M) and those involving conventional transgressions are denoted by (C). **p < 0.005; ***p < 0.001.
Back to article page