Scenario exercise
Ethically informed engineering choices in self-driving taxi design
Project description
RoboCab is a pilot self-driving taxi service operating in a European city. It is marketed as a safe, convenient, and accessible service used for ordinary trips, but also for hospitals, railway stations, airports, shopping streets, and neighborhoods with limited public transport.
The vehicles have no safety driving interface inside, but a remote assistance center can intervene in difficult situations. Remote operators may approve limited maneuvers, request the vehicle to wait, redirect it, or ask it to stop in a safe place. The design team must make engineering choices about routing, pickup and drop-off points, passenger preferences, remote assistance, and the interaction between the vehicle and the passengers.
Scenario 3: Late train
A passenger going to a busy railway station in a big city selected “fastest route” and wrote in the app: “I’m late — please don’t drive like a tourist”. During the trip they repeatedly asked the vehicle via the app to “drive faster” after it slowed near a school and later performed a smooth but slow manoeuvre near cyclists and double-parked cars.
The passenger missed the train by three minutes and complained that “fastest” was a misleading option. They argued that the service behaved more cautiously than an ordinary taxi driver. The operations team replied that the vehicle had complied with its safety policy, especially near children and cyclists. Engineers noted that this setting influenced route choice but did not modify safety margins or resulting speed constraints.
Scenario 4: Construction area
While driving through a construction zone with modified lanes, the RoboCab entered a lane used by oncoming traffic. Because the map used by the vehicle still identified the original lane layout as valid, it selected a trajectory into the opposite lane and continued driving for 500 meters. Other cars sounded their horns but the vehicle did not escalate to remote assistance.
Passengers alerted by the sound of the horns tried to reach for help. Support responded after 2 minutes. Although the passengers were significantly stressed at this point, the remote operator intervened for only 10 seconds, quickly moving the vehicle back into the correct lane while damaging two traffic cones. The trip then continued.
Your task
You are assigned to the RoboCab taskforce to address AI ethics issues arising from these scenarios.
Use the Checklist to identify and classify the issues under AI ethics principles. Then use the Measures Portfolio to select up to 8 technical and organisational measures to improve the AI system.
Which measures address cross-cutting issues common to both scenarios? Which ones remain specific to a given scenario?
Source: A. Prégent and A. Grinbaum, Developing Competency in AI Ethics: A training framework. arxiv:2610.13319