Happy Driver

An online experiment investigating whether passenger mood affects preference for driving style in self-driving cars. 182 participants were assigned to one of three induced mood conditions — calm, neutral, excited — and rated three driving styles: conservative, moderate, and aggressive.

The main finding: aggressive driving was universally disliked, regardless of mood. The more nuanced result: a mismatch between mood and driving style predicted lower satisfaction more strongly than either variable alone. A calm passenger in an aggressively-driven car was more dissatisfied than the raw driving style rating predicted. The AI's driving style and the passenger's emotional state needed to be compatible, not just independently acceptable.

The design implication: self-driving systems should not simply optimise for the driving style passengers rate highest in isolation. They should adapt to the passenger's current state. Personalisation in human-AI systems often means adapting to stable user preferences; this work suggests it also needs to adapt to dynamic user states.

Read the paper →