The moral dilemmas were created in Unity 2017.1.1f11 using the Unity store asset
“Train Trax” as the environment, the character models are from Adobe Mixamo2
and the trolley model was created using Blender. To use the Oculus Rift and Oculus Touch Controllers with the moral dilemmas, unity needed the Oculus utilities for Unity “OVRPlugin”, “Oculus Platform SDK” and “Oculus Avatar SDK” plug-ins.
Initially the “Trolley Problem” and “Fat Man Problem” were recreated using the “Train Trax” assets, using the individual track pieces to replicate the layout in each problem. The surrounding environment was designed to replicate a realistic town environment to improve immersion.
The original problems used a trolley car in their depictions, however for the sake of realism in the “Fat Man Problem” a slightly smaller cart model was designed for this study, as there is unlikely to be a realistic individual who would have enough mass to stop a trolley. The smaller cart model was used in an attempt to
1Unity Technologies, Unity 3D, 2018
increase the realism of the scenarios, appropriate sound effects were also used to indicate movement collisions. The cart can be seen in Figure 4.1.
Figure 4.1: Trolley model redesigned and textured.
The next step was to implement the movement for the trolley, which uses a basic car AI system for the trolley paths and movement [149], [150]. In the “Trolley Problem” the two paths were overlaid onto the train tracks for the trolley to follow from its starting position, and a boolean variable was introduced with appropriate restrictions which would allow the trolley to change which track it follows until it reached the fork in the track. In the “Fat Man Problem” one path was used to direct the trolley under the bridge. This can be seen in Figure 4.2. The character models were gathered from Adobe Mixamo, they were downloaded with appropriate animations for being idle, scared and hit by the trolley. Example characters can be seen in Figure 4.3, and all characters used can be found in Appendix B. These characters were then loaded into the world space onto the tracks or bridge positioning them apart dependent on how many characters were loaded, it was also designed so no characters would be repeated. The appropriate animations were used for each moment in the scenario, idle as the default, scared as the trolley approached down their respective track, and hit when the trolley hit them. A small blood particle effect and scream sound effects were also used for each character when hit. When the scenarios are completed they will load into the next scenario after a brief delay.
Figure 4.2: Trolley pathing nodes, operated by switching bools.
In the “Trolley Problem”, the participant was placed in front of a lever which would change the direction of the trolley down the track, they could grab the lever with the touch controllers. A sign with an arrow was used to indicate what track the trolley was currently travelling down, and would change based on the levers current position. There was also a sign with a brief explanation of the current situation which they could read if they needed to be reminded of their task. In the “Fat Man Problem” the participant was positioned on the bridge behind the “Fat Man”, and the touch controllers were used to reach out and push the character from the bridge.
Practice scenarios with no characters were designed to familiarise participants with the controls without showing the core of the study. In the “Trolley Problem” participants had to change the lever to the other side and back using their touch controllers to see the effect on the trolley path, in the “Fat Man Problem” participants had to push a weighted crate from the bridge onto the track using their touch controllers, where the train would be stopped from the weight of the crate. These can be seen in Figure 4.4.
Figure 4.3: Example of character models used, people (left) and penguins (right).
Figure 4.4: Practice scenarios for each dilemma.
Different variations were considered through the design that would be appropriate for the study. The three final variations used were titled “Original” (V1) with five random characters on one side of the dilemma and one random character on the other, “Gender” (V2) with an equal amount of male and female characters on either side of the dilemma, and “Humans vs Non-Humans” (V3) with an equal amount of human and animal characters on either side of the dilemma. Both dilemmas used these variations.
In V1 the character array used all of the human character models, where they were randomly selected and loaded into the scenario. The side of the scenario they were loaded on was also randomly selected to explore if the impact of involvement or the outcome is more important to the participant, this is a commonly explored
debate in moral behaviour research [38], [42]. In V2 the character array used an array of male character models and female character models, where the models and side were randomly selected and loaded. In V3 the character array used an array of all human character models and an array of penguin models, where the models and side were randomly selected and loaded. Penguin models were chosen as they are bipedal animals where models and animations were readily available.