Human-Centric explanations for users in automated Vehicles: A systematic review.

Area of Science:

• Human-Computer Interaction
• Automated Vehicle Systems
• Trustworthy AI

Background:

• Automated vehicle (AV) decision-making can reduce user trust and acceptance.
• Human-centric explanations are crucial for transparency, trust, and safe human-vehicle interaction.
• Limited research exists on how explanation types affect drivers and how to evaluate them.

Purpose of the Study:

• To systematically review human-centric explanations in automated vehicle contexts.
• To identify best practices for what, when, and how to provide explanations.
• To address the impact of explanations on user trust, acceptance, and situational awareness.

Main Methods:

• Systematic literature review following PRISMA guidelines.
• Searched five major databases (Scopus, Web of Science, IEEE Xplore, TRID, Semantic Scholar) from 2000 to April 2024.
• Included 59 articles out of 266 identified.

Main Results:

• Explanations clarifying reasoning are more effective than those describing actions.
• Explanations before action are recommended, but optimal timing requires further study.
• Multimodal explanations are best when distinct; otherwise, visual-only is preferred. Narrative perspective impacts trust differently.

Conclusions:

• Human-centric explanations must be tailored to specific driving contexts.
• Future research should focus on explanation length, timing, modality, and real-world validation.
• Optimized explanations are vital for safer, more trustworthy human-AV interactions.