Do Autonomous Vehicles Dream of Tort Liability?

Kyle Moodhe, Contributing Member 2023-2024
Intellectual Property and Computer Law Journal

I. Introduction

The technological leaps of artificial intelligence (“AI”) and machine learning in recent years have been nothing short of astounding. The future appears bright for the use of AI and autonomous machines in everyday life.[1] However, autonomous machines, like humans, are not free from all error. When these errors lead to harm, how does tort law treat these computer tortfeasors?

Many legal scholars of autonomous vehicles believe that one day the safety of autonomous vehicles will surpass the safety of vehicles operated by humans.[2] However, some scholars believe manufacturers will resist adopting safer autonomous vehicles because tort law, as currently constructed, disfavors autonomous vehicles by imposing the costly standards of strict liability.[3] Some scholars argue that in order to incentivize the adoption of safer autonomous vehicles, tort law should move towards negligence and away from strict liability.[4] In other words, strict liability is too burdensome on autonomous vehicle manufactures, and it prevents long-term safety gains that occur when human error is no longer a possibility. The bargain is to sacrifice the small increases in safety in the short-term for a greater net gain in safety through incentivizing widespread adoption of autonomous vehicles in the long-term.[5]

Other scholars argue that the traditional tort law system, including product liability, is capable of adjudicating liability for autonomous vehicle crashes.[6] One scholar suggests that concerns about hindering the adoption of autonomous vehicles should be “irrelevant in evaluating [manufacturer] liability in tort.”[7]

This article presents two scholarly visions of allocating tort liability in autonomous vehicle accidents and evaluates the strengths and weaknesses of each vision.

II. Background

Automated vehicles, specifically cars, exist along a spectrum of human control.[8] In fact, the Society of Automotive Engineers (SAE) introduced a six-level system of classification, which ranges from complete human control at level 0 to complete computer control at level 5.[9] In between these levels exist varying degrees of human control and automated features.[10] Anti-lock braking systems, electronic stability control, and automatic emergency braking are some of the widespread and existing features of cars that could be considered autonomous.[11] This means most cars currently fit in the middle levels of SAE’s classification system.[12]

Car crashes are the leading cause of death for young adults ages 18-24, and an overwhelming 94% can be attributed to human error.[13] Fully autonomous vehicles are not subject to common causes of car fatalities due to human error such as impairment, distraction, or sleep deprivation.[14] Although fully autonomous vehicles are not commercially available, current vehicles with autonomous features have already improved vehicle safety.[15] For example, the National Transportation Safety Board (NTSB) estimates 80% of all rear-end collision fatalities could be eliminated with the universal implementation of automatic emergency braking.[16] Other organizations predict that once autonomous vehicles can prove ten times safer than human drivers, they will reduce the number of car accident fatalities to less than 4,000 deaths.[17] For reference, in 2022, the National Highway Traffic Safety Administration (NHTSA) reported 42,795 car accident fatalities in the United States.[18] Most scholars in this field believe that it is not a matter of if automated vehicles will surpass driving ability but of when.[19]

However, there are many compelling reasons to reject the notion that autonomous vehicles will surpass human drivers in terms of safety. First, human drivers are relatively safe drivers, statistically speaking.[20] In fact, humans do not crash for 99.999819% of miles they travel.[21] Put another way, humans have 1.81 crashes for every 1 million miles traveled.[22] Humans in the United States travel, by car, roughly 3 trillion miles per year.[23] Some studies show that autonomous vehicles would need to travel many millions of miles more before any accurate statistical comparison could be made.[24] Further skewing our knowledge of the performance of autonomous vehicles is the fact that current semi-autonomous vehicles allow for computer-to-human handoffs of driving control.[25] Semi-autonomous vehicles may experience sensory issues during times of pouring rain and handoff control to the human in the driver seat.[26] As a result, we do not have enough information demonstrating the performance of autonomous vehicles in sensorially complex situations.

III. Discussion

Autonomous vehicles offer new conceptual legal challenges with the introduction of a computer tortfeasor. One of the interesting challenges that autonomous vehicles present is that a computer tortfeasor cannot be meaningfully punished for its actions.[27] Additionally, there is no way in which computer tortfeasors can be financially deterred from repeating unreasonable conduct.[28] Lastly, it may be impossible to ever understand the thought process of a computer tortfeasor whose decision-making is somewhat of a black box.[29] In spite of these differences from human tortfeasors, there may be traditional tort law frameworks which can encompass the problems posed by autonomous vehicles.[30] Some scholars believe that autonomous vehicles can operate under the strict liability framework applied to animal tortfeasors.[31] Others have suggested that agency rules of vicarious liability are a sufficient framework for allocating liability between a computer tortfeasor as agent and the manufacturer as principal.[32]

One scholar, Ryan Abbott, argues that as autonomous vehicles become safer, we ought to adjudicate autonomous vehicle accidents under negligence and adopt a new standard of care in order to encourage more widespread adoption of autonomous vehicles.[33] Abbot argues that tort law has the potential to incentivize or disincentivize the adoption of safer new technologies.[34] Abbott argues that negligence adeptly balances the interest of promoting safety, compensating victims, and encouraging economic growth.[35] The reasonability of the tortfeasor’s actions determines whether or not the tortfeasor is liable for any harm caused.[36] Abbot argues that tort law should incentivize safety by deterring unsafe behaviors.[37] Negligence should only be adopted for computer tortfeasors once it can be demonstrated that those computer tortfeasors are reliably safer than their human counterparts.[38]

The alternative to evaluating autonomous vehicle tort liability under negligence is by a no-fault theory of strict liability used in cases of product liability.[39] Strict liability allocates liability to the tortfeasor regardless of whether or not the conduct was wrong or blameworthy.[40] Product liability requires that the plaintiff demonstrate the manufacturer’s product was defective, flawed, or incorrectly advertised.[41] Currently, human tortfeasors are judged under the standards of negligence, whereas computers fall under the harsher scheme of strict liability.[42] Thus, computer tortfeasors are held to a higher standard, even if they are safer than human tortfeasors, who may be more unsafe.[43]

Abbott argues that computer tortfeasors are no longer mere products.[44] Instead, they more closely resemble human tortfeasors, so they should be held liable under negligence rather than products liability.[45] Abbott goes further and argues that at the point when computers perform tasks more safely than humans, then humans should be held to a “reasonable computer” standard of care.[46] This reasonable computer standard would create liability for humans who engage in tasks that were safer to perform by computers.[47]

Alexander Lemann believes that existing tort law paradigms are sufficient for handling the problems posed by autonomous vehicles.[48] Lemann believes that the utilitarian and economic policies underlying arguments for changing tort law to accommodate autonomous vehicles are irrelevant considerations.[49] Lemann points to the very fact that tort law exists to provide a remedy not just for injury but for the wrong of an injury as well.[50] Although the consumer expectations test has been largely abandoned, the moral priority of protecting consumers from defects they did not knowingly choose to expose themselves underlies the law of products liability to this day.[51] For Lemann, the history of products liability for automobile manufacturers proves that tort law can handle autonomous vehicles without a major revolution in the law.[52]

Lemann believes that autonomous vehicles must be held to the products liability standards that all vehicles are subject to, irrespective of their safety in relation to conventional vehicles.[53] He asserts, “Just as a manufacturer cannot defend an allegation of design defect by pointing out that its car is safer than a horse, so an autonomous vehicle manufacturer should not be able to point to a conventional vehicle in defending itself.”[54] The unique problem of autonomous vehicles is the significant unknowability of the decision-making process that leads to flawed results.[55] However, traditional tort concepts like res ipsa loquitur (that the existence of negligence “speaks for itself” based on the circumstances of an accident or harm) can still be employed in these edge cases of inexplicable artificial intelligence malfunction.[56] The more one probes tort law, Lemann argues, the less and less unqualified it appears to be for the task of handling autonomous vehicles.[57] Products liability, in particular, appears adequate to the task of applying old concepts to new problems, even autonomous vehicles.

IV. Conclusion

Lemann’s argument for tort law continuity is more compelling than Abbott’s desire to create a new tort scheme for autonomous vehicles. One of the main critiques of Abbott’s argument is how much it relies on the premise that autonomous vehicles will perform magnitudes more safely than human drivers. However, there are deep legal and philosophical questions that arise with a computer tortfeasor that cannot be easily discarded. While significant changes in the law may result as society grapples with the unknowable decision-making processes of autonomous vehicles, it may be short-sighted to abandon flexible tort doctrines for a novel legal regime.

[1] Adrian Bridgwater, What Drove the AI Renaissance?, Forbes (Apr. 15, 2019, 4:20 AM), https://www.forbes.com/sites/adrianbridgwater/2019/04/15/what-drove-the-ai-renaissance/?sh=2badb9d1f25c

[2] Alexander B. Lemann, Autonomous Vehicles, Technological Progress, and the Scope Problem in Products Liability, 12 J. Tort L. 157, 159 (2019)

[3] Ryan Abbott, The Reasonable Computer: Disrupting the Paradigm of Tort Liability, 86 Geo. Wash. L. Rev. 1, 21-22 (2018)

[4] Id. at 22.

[5] Id. at 22-23

[6] Lenmann, supra note 2 at 159.

[7] Id.

[8] Abbott, supra note 3 at 30.

[9] Joseph Avery, Fumble! Anti-Human Bias in the Wake of Socio-Technical System Failures, 53 Ariz. St. L.J. 1009, 1027 (2021)

[10] Abbott, supra note 3 at 30.

[11] Lenmann, supra note 2 at 164.

[12] Avery, supra note 9 at 1027.

[13] Lenmann, supra note 2at 158.

[14] Id. at 172-173.

[15] Id. at 174.

[16] Id. at 174-175.

[17] Abbott, supra note 3 at 18-19.

[18] US traffic deaths drop slightly in 2022 but still a ‘crisis’, Associated Press (Apr. 20, 2023, 5:55 PM), https://apnews.com/article/traffic-deaths-distracted-driving-crisis-6db6471e273b275920b6c4f9eb7e493b

[19] Lenmann, supra note 2 at 158.

[20] Steve Silva, Human Drivers Avoid Crashes 99.999819% of the Time, Self-Driving Cars Need to Be Even Safer, Jalopnik (Mar. 3, 2023), https://jalopnik.com/self-driving-car-vs-human-99-percent-safe-crash-data-1850170268

[21] Id.

[22] Id.

[23] Lenmann, supra note 2 at 174.

[24] Id.

[25] Avery, supra note 9 at 1009.

[26] Id. at 1023.

[27] Karni A. Chagal-Feferkorn, How Can I Tell If My Algorithm Was Reasonable?, 27 Mich. Tech. L. Rev. 213, 249 (2021)

[28] Id. at 249.

[29] Id. at 234-35.

[30] Lenmann, supra note 2 at 180.

[31] Id.

[32] Chagal-Feferkorn, supra note 26 at 249.