Fig 1.
Example of an illustration of an accident scenario.
In this example, the passenger-to-pedestrian ratio is 1:2, which means that the life of one passenger is weighed against that of two pedestrians. The visual illustrations of the scenarios were created using Microsoft PowerPoint.
Fig 2.
The multinomial processing tree model used in Experiments 1a and 1b.
The rectangles on the right represent the answer categories available in each condition. Parameter πDQ represents the parameter estimate for the preference that the autonomous vehicle should sacrifice the passenger instead of the pedestrian/s. Separate model trees were necessary for each combination of the 3 (perspective: pedestrian, observer, passenger) × 4 (passenger-to-pedestrian ratio: 1:1, 1:2, 1:5, 1:10) design. Note that the model corresponds to the model representing the direct questioning approach in Experiment 3.
Fig 3.
Descriptive data for Experiment 1a.
The probability of sacrificing the passenger rather than the pedestrian/s is depicted as a function of passenger-to-pedestrian ratio (1:1, 1:2, 1:5, and 1:10) and perspective (pedestrian, observer, and passenger). The error bars represent bootstrapped standard errors.
Table 1.
Comparisons among perspectives separately for each passenger-to-pedestrian ratio in Experiment 1a.
Fig 4.
Descriptive data for Experiment 1b.
The probability of sacrificing the passenger rather than the pedestrian/s is depicted as a function of passenger-to-pedestrian ratio (1:1, 1:2, 1:5, and 1:10) and perspective (pedestrian, observer, and passenger). The error bars represent bootstrapped standard errors.
Table 2.
Comparisons among perspectives separately for each passenger-to-pedestrian ratio in Experiment 1b.
Fig 5.
Descriptive data for Experiment 2a.
The probability of sacrificing the passenger/s rather than the pedestrian is depicted as a function of passenger-to-pedestrian ratio (1:1, 2:1, 5:1, and 10:1) and perspective (pedestrian, observer, and passenger). The error bars represent bootstrapped standard errors.
Table 3.
Comparisons among perspectives separately for each passenger-to-pedestrian ratio in Experiment 2a.
Fig 6.
Descriptive data for Experiment 2b.
The probability of sacrificing the passenger/s rather than the pedestrian is depicted as a function of passenger-to-pedestrian ratio (1:1, 2:1, 5:1, and 10:1) and perspective (pedestrian, observer, and passenger). The error bars represent bootstrapped standard errors.
Table 4.
Comparisons among perspectives separately for each passenger-to-pedestrian ratio in Experiment 2b.
Fig 7.
The illustration that was used for the moral dilemma scenario.
The passenger-to-pedestrian ratio was 5:1, which implies that the life of five passengers was weighed against that of one pedestrian. The visual illustration of the scenario were created using Microsoft PowerPoint.
Fig 8.
Multinomial processing tree model.
The combined multinomial processing tree model for the Direct Questioning Group—represented by the upper tree—and for Indirect Questioning Groups 1 and 2—represented by the lower two trees—for the Extended Crosswise Model [45] adapted to the present experiment. The rectangles on the right contain the answer categories available in each condition. Parameter π represents the prevalence estimates for the preferences that the autonomous vehicle should sacrifice five passengers of the autonomous vehicle in order to save the pedestrian, depending on the condition. Parameter pNov-Dec represents the known prevalence of the non-sensitive attribute, in this case, the participant being born in November or December.