Actuarial Insights on the Risks of Tomorrow Autonomous Vehicles Gamma Iota Sigma Webinar Series April 12, 2017 Rick Gorvett, FCAS, CERA, MAAA, ARM, FRM, PhD Staff Actuary Casualty Actuarial Society
2
Agenda Background Issues Opportunities 3
1964 Worlds Fair General Motors Futurama Knight Rider (with KITT) 2015 Mercedes Concept Car Google Self-Driving Car/Taxi 4
Historic Development 2013 - Google surpasses 500K miles - Oxford creates a $7,750 self-driving car - Britain tests on public roads - Mercedes tests on public roads - CMU tests on public roads - Audi receives autonomous car license - NHTSA issues policy on automated vehicles - DC passes autonomous car law 2011 - Google surpasses 150K miles - BMW begins testing self driving car on public roads - NV passes autonomous car law 2010 Volvo CitySafe standard 2014 - MI passes law - NHTSA passes V2V - Google developing driverless car without steering wheel or brakes 2012 - Google surpasses 300K accident free miles - Nissan opens research facility in Silicon Valley - Google & Continental receive autonomous car licenses - FL & CA pass autonomous car laws 2009 - Google begins testing on public roads 2007 CMU wins DARPA Urban Challenge 2005 Stanford wins DARPA Grand Challenge 5
Even More Recently The Good Self-Driving Buses Japan Helsinki Self-Driving Taxis The Bad Accidents Google caused accident Tesla fatal crash The Ugly Some AVs are not exactly stylish 6
Level of Vehicle Automation Autonomous Vehicles (AV): Vehicles that are able to guide themselves from an origin point to a destination point desired by the individual Varying levels of Automation (by NHTSA): Level 0 No Automation Level 1 Function- Specific Automation (e.g. cruise control) Level 2 Combined Function Automation (e.g. adaptive cruise control with lane centering) Level 3 Limited Self- Driving Automation (e.g. drivers can cede safety-critical functions) Level 4 Full Self- Driving Automation 7
Enabled by Connected Vehicles LIDAR: combination of light and radar, and uses laser light to create 3D images of the surrounding environment. Video Camera Ultrasonic Sensor Computer RADAR V2V/V2I uses Dedicated Short Range Communications (DSRC), similar to wifi 8
Future development may create two models for AVs All driving, limited location Some driving, all locations End to end service Only operates in specified area Taxi service Google, Uber Takes over some of the driving E.g. Supercruise, parallel parking Only operates in specified area Driver owns and operates Mercedes, BMW, Volvo, Cadillac, Telsa
Societal Benefits of AV Reduce accidents Reduce transportation costs Support demographic change Promote the economy 10
Agenda Background Issues Opportunities 11
CAS AVTF: Overview Goal The CAS AVTF is researching the technology s risks to provide policymakers with the information needed to ensure the product is brought to market as safely and efficiently as possible. Focus Pre market Post market Post claim identify & quantify risks accurately price the technology compensate claimants fairly & efficiently Taskforce is actively pursing relevant studies and other opportunities 12
Issues 1. Safety Are these vehicles safe? What should the safety standard be? 2. Liability Who is liable in the event of an accident? 3. Regulation What regulations should govern the testing and driving of an AV? 4. Privacy and Cyber Security Who owns and is responsible for the data collected by AVs 13
93% of accidents are caused by human error. NHTSA s 2008 National Motor Vehicle Crash Causation Survey Automated vehicles will reduce accidents by 93% 14
NMVCCS Limiting Factors 50% Technology Issues Behavioral (Driver) Issues 48.9% 40% 30% 20% 10% 0% 32.4% 21.3% 16.7% 12.2% 11.6% 11.0% 0.4% 3.1% 2.3% 2.9% 1 2 3 4 1 2 3 4 5 6 15
Some Automated Vehicle Caveats 16
NMVCCS Implications of the CAS Study New benchmark should be calculated Data is old and unrepresentative Human driving risks automated vehicle risks Appropriate test for each risk Computer simulations for technology s error rate Simulations provide little insight into driver s actual use of technology. Policy changes can increase AV s safety 1% reduction in accidents is ~55k fewer accidents and $1.4 billion of economic value per year Policy cost benefit analysis E.g. driver training program, automated vehicle only lanes, allowing the AVs to speed 17
Actuarial Pricing of Auto Insurance Cost-Based pricing approach As auto insurance losses decrease, premiums eventually decrease As opposed to a Market-Based pricing approach Charge what the market allows Law of large numbers Risks grouped by characteristics Rates charged based on group rating Actual discount determined by vehicle rating Rating Characteristic Examples Driver age Location Driving history Mileage Vehicle
Types of Auto Coverage First-Party Liability Comprehensive: Expenses due to theft, vandalism, glass breakage, and related matters to your car that weren't caused by an auto accident. Collision: Damages incurred by your vehicle in an auto accident. Medical payment coverage: Cover medical expenses you incur up to a limit Uninsured/underinsured motorist: Cover Others: Towing/Rental Bodily Injury: Medical-related expenses you caused to others. Physical damage: Cost to repair or replace other's property (such as a car)? Coverage not as affected in a world of AVs 19
Possible Insurance Frameworks for AVs 1. Product Liability Attach liability to sellers and manufacturers of the vehicle Tends to be complex and expensive as the standard to establish a defect is vague/unpredictable 2. Strict liability when an AV is at fault Making the owner of the vehicle responsible when the owner s automobile is at fault 3. First party insurance Similar to UM coverage, injured parties would look to their own insurers 4. A combination of above? 20
Current U.S. regulatory approach varies by state http://cyberlaw.stanford.edu/wiki/index.php/automated_driving:_legislative_and_regulatory_action 21
Agenda Background Issues Opportunities 22
Insurance Industry Will Add Value More detailed accident data & models Risk management expertise Best understanding of 51 different state driving regulations Best understanding of products liability & general liability Financial incentive to decrease losses A commitment to charge rates that are not excessive, inadequate or unfairly discriminatory 23
Actuarial Opportunities Responsible for matching price to risk Past Future: Represents a fundamental change in relationship between driver & vehicle Heterogeneous: Different products perform differently Black Box: Cannot readily discern differences Outside influence: Outside interests may put pressure on rates Big Data 1 GB/s data generated Machine/Deep Learning 24
Other Considerations Adoption of AVs New type of transportation? Replacement car? Infrastructure Planning Car Ownership Pattern Traveler Behavior Pattern Impact on public transportation? 25
Speaking of Behavior Structuring publicly-acceptable algorithms Philosophical / ethical / social Philippa foot (trolley problem) Hobbes, Locke, Rousseau (social contract) Omission vs commission Individual vs public With what rules do we program Avs? moralmachine.mit.edu 26
You re Invited join online today at www.casstudentcentral.org
CAS Student Central Membership program for university students: No Membership fee Access to resources including P&C internship listings, CAS Curriculum Guide, Case Studies, Online Community Free webinars created specifically for students Invitations to free networking events including student programs at CAS meetings and Seminars Thousands of student members from hundreds of schools
29
Questions and Discussion rgorvett@casact.org