Establishing whether automated vehicles (AVs) are acceptably safe is not straightforward, and continual technology modification adds complication. RAND researchers analyzed three approaches to assessing safety—measurements, processes, and thresholds—and noted the different kinds of evidence associated with each. This report also describes the challenges and options related to communicating about AV safety, especially to the general public.

Cover: Safe Enough
Read Online

Safe Enough

Approaches to Assessing Acceptable Safety for Automated Vehicles

by Marjory S. Blumenthal, Laura Fraade-Blanar, Ryan Best, J. Luke Irwin

Download

Download eBook for Free

FormatFile SizeNotes
PDF file 1.5 MB

Use Adobe Acrobat Reader version 10 or higher for the best experience.

Purchase

Purchase Print Copy

 FormatList Price Price
Add to Cart Paperback142 pages $39.00 $31.20 20% Web Discount

Research Questions

  1. In what ways can AV safety can be assessed?
  2. How can measurements be used to assess AV safety?
  3. How can processes be used to assess AV safety?
  4. How can thresholds be used to assess AV safety?
  5. How do different approaches to assessing AV safety compare and interact with each other?
  6. How can communication about AV safety be improved?

Establishing whether automated vehicles (AVs) are acceptably safe is not straightforward, and continual technology modification adds complication. RAND Corporation researchers analyzed three categories of approach—measurements, processes, and thresholds—and noted the different kinds of evidence associated with each, the ways in which different approaches can be used together, and the degree to which stakeholder groups agree on the merits of these approaches. This report complements discussion of measurement and analytical issues with a discussion of challenges in communicating about AV safety, especially to the general public. Its recommendations are aimed at both industry and government.

Key Findings

Different approaches to assessing AV safety—measurement, process, and thresholds—complement each other, with no one best approach

  • Leading measures remain key to assessing AV safety given the lack of crash-based lagging measures.
  • Among leading measures, approaches to roadmanship, the ability to drive on the road safely without creating hazards and responding well to hazards created by others, are growing.

Because existing measures are insufficient to assess AV safety, more attention is being paid to processes—steps taken by developers and how these steps are implemented—and what they indicate

  • These processes include safety-relevant standards-setting activities, a growing emphasis on safety culture, and widening use of safety cases (collections of assertions defining how safety is being promoted and assessed).
  • Compliance with regulations is also a process, but its reach remains smaller than that for conventional vehicles.

Thresholds—which can be qualitative as well as quantitative—are another way to assess AV safety

  • The most straightforward threshold is human driving performance.
  • Other thresholds relate to AV performance and absolute goals.
  • Thresholds will evolve as technology develops, commercial use expands, and expectations rise. Meeting thresholds is not a one-time achievement.

Communicating about AV safety shapes public trust

  • Although messages about AV safety come from many sources, the provenance of messages is not always obvious to the general public.
  • A RAND American Life Panel survey demonstrates the importance of how communications with the general public are structured.
  • The public has the greatest trust in AV safety messages from government.

Recommendations

  • AV developers and the larger AV research community should continue to advance and integrate leading measures, including roadmanship. An industry consortium could help.
  • Developers should forge uniform and transparent approaches to presenting evidence for meeting safety thresholds. These would factor in differences in use cases, operational design domains, and more, and they could focus on better-than-average driving as a threshold.
  • The U.S. Department of Transportation should support further research into human drivers to enable the comparisons that stakeholders seek. Such research should address differences in operational design domains.
  • AV developers should collaborate with state and local leaders to bring their vehicles into communities around the country. Direct exposure to AVs supports public understanding.

Table of Contents

  • Chapter One

    Introduction

  • Chapter Two

    Safety as a Codified Concept for AVs

  • Chapter Three

    Safety as a Measurement

  • Chapter Four

    Safety as a Process

  • Chapter Five

    Safety as a Threshold

  • Chapter Six

    Communicating About Safety

  • Chapter Seven

    Conclusions and Recommendations

  • Appendix A

    American Life Panel Survey

  • Appendix B

    Interviews

  • Appendix C

    Literature Highlights

Research conducted by

This research was commissioned by the Uber Advanced Technologies Group and conducted in the Community Health and Environmental Policy Program within RAND Social and Economic Well-Being.

This report is part of the RAND Corporation Research report series. RAND reports present research findings and objective analysis that address the challenges facing the public and private sectors. All RAND reports undergo rigorous peer review to ensure high standards for research quality and objectivity.

This document and trademark(s) contained herein are protected by law. This representation of RAND intellectual property is provided for noncommercial use only. Unauthorized posting of this publication online is prohibited; linking directly to this product page is encouraged. Permission is required from RAND to reproduce, or reuse in another form, any of its research documents for commercial purposes. For information on reprint and reuse permissions, please visit www.rand.org/pubs/permissions.

The RAND Corporation is a nonprofit institution that helps improve policy and decisionmaking through research and analysis. RAND's publications do not necessarily reflect the opinions of its research clients and sponsors.