Safer than the average human driver (who is less safe than me)? Examining a popular safety benchmark for self-driving cars

Inter-reviewer reliability of human literature reviewing and implications for the introduction of machine-assisted systematic reviews: a mixed-methods review

OBJECTIVES

Our main objective is to assess the inter-reviewer reliability (IRR) reported in published systematic literature reviews (SLRs). Our secondary objective is to determine the expected IRR by authors of SLRs for both human and machine-assisted reviews.

METHODS

We performed a review of SLRs of randomised controlled trials using the PubMed and Embase databases. Data were extracted on IRR by means of Cohen's kappa score of abstract/title screening, full-text screening and data extraction in combination with review team size, items screened and the quality of the review was assessed with the A MeaSurement Tool to Assess systematic Reviews 2. In addition, we performed a survey of authors of SLRs on their expectations of machine learning automation and human performed IRR in SLRs.

RESULTS

After removal of duplicates, 836 articles were screened for abstract, and 413 were screened full text. In total, 45 eligible articles were included. The average Cohen's kappa score reported was 0.82 (SD=0.11, n=12) for abstract screening, 0.77 (SD=0.18, n=14) for full-text screening, 0.86 (SD=0.07, n=15) for the whole screening process and 0.88 (SD=0.08, n=16) for data extraction. No association was observed between the IRR reported and review team size, items screened and quality of the SLR. The survey (n=37) showed overlapping expected Cohen's kappa values ranging between approximately 0.6-0.9 for either human or machine learning-assisted SLRs. No trend was observed between reviewer experience and expected IRR. Authors expect a higher-than-average IRR for machine learning-assisted SLR compared with human based SLR in both screening and data extraction.

CONCLUSION

Currently, it is not common to report on IRR in the scientific literature for either human and machine learning-assisted SLRs. This mixed-methods review gives first guidance on the human IRR benchmark, which could be used as a minimal threshold for IRR in machine learning-assisted SLRs.

PROSPERO REGISTRATION NUMBER

CRD42023386706.

"It's okay because I'm just driving": an exploration of self-reported mobile phone use among Mexican drivers

Introduction

Technological advancements have the potential to enhance people's quality of life, but their misuse can have a detrimental impact on safety. A notable example is the escalating issue of distracted driving resulting from the use of mobile phones behind the wheel, leading to severe crashes and injuries. Despite these concerns, both drivers' usage patterns and their risk-related associations remain scarcely documented in Mexico. Therefore, this descriptive study aimed to examine the mobile phone usage of Mexican drivers, its relationships to risk awareness and near-miss/crash involvement, and the self-reported underlying reasons for this behavior.

Methods

This cross-sectional study utilized a sample of 1,353 licensed Mexican drivers who took part in a nationwide series of interviews regarding their onboard phone use settings.

Results

A significant percentage of drivers (96.8%) recognize using a mobile phone while driving as high-risk behavior. However, only 7.4% reported completely avoiding its use while driving, with 22.4% identified as high-frequency users. Frequency was also found positively associated with the self-reported rate of near-misses and crashes. Furthermore, qualitative data analysis highlights the emergence of a 'sense of urgency' to attend to phone-related tasks in response to daily demands and life dynamics, offering a potential explanation for this behavior.

Conclusion

The results of this study suggest common patterns of onboard mobile use among Mexican drivers concerning driving situations and associated risks. This underscores the need for increased efforts to discourage onboard phone use in the country.

Is safety in the eye of the beholder? Discrepancies between self-reported and proxied data on road safety behaviors—A systematic review

Recent studies have problematized on the lack of agreement between self-reported and proxied data in the field of road safety-related behaviors. Overall, and although these studies are still scarce, most of them suggest that the way we perceive our own road behavior is systematically different from the perspective from which we perceive others' behavior, and vice versa. The aim of this review paper was to target the number and type of studies that have researched the behavioral perceptions of different groups of road users, contrasting self-reported behavioral data with those reported by other users (proxied), and their outcomes. This systematic review followed the PRISMA methodology, which allows for the identification of relevant articles based on the research term. A total number of 222 indexed articles were filtered, and a final selection of 19 articles directly addressing the issue was obtained. Search strategies were developed and conducted in MEDLINE, WOS, Scopus and APA databases. It is remarkable how road users perceive themselves as behaviorally “safer” than the rest of road users in what concerns the knowledge of traffic norms and their on-road performance. In addition, and regardless of the type of user used as a source, self-reported data suggest their perceived likelihood to suffer a traffic crash is lesser if compared to any other user. On the other hand, proxied reports tend to undervalue third users' performance, and to perceive riskier behaviors and crash-related risks among them. The outputs of this systematic review support the idea that the perception of road users' behavior and its related risks substantially differ according to the source. It is also necessary to increase the number, coverage and rigor of studies on this matter, perhaps through complementary and mixed measures, in order to properly understand and face the bias on road users' risk-related behaviors.

Is it OK to bully automated cars?

To integrate automated vehicles (AVs) into our transportation network, we should consider how human road users will interact with them. Human aggression toward AVs could be a new risk in mixed traffic and reduce AV adoption. Is it OK to drive aggressively toward AVs? We examined how identical aggressive behavior toward an AV or human driver is appraised differently by observers. In our 2 (scenario type: human driver vs. AV) × 2 (victim identity salience: low vs. high) between-subjects survey, we randomly allocated participants (N = 956) to one of four conditions where they viewed a video clip from an AV or a human driver showing a car suddenly braking continuously ahead of the AV or human driver's car. The salience of victim identity influenced the observers' appraisals of aggressive behavior. When asked to judge the front car's behavior toward this AV or human driver (the victim identity is salient), they reported more acceptability and less risk perception, negative affect, and immoral judgment while judging this behavior toward the AV. When asked to judge the front car's behavior (the victim identity not highlighted), they reported non-different appraisals. This finding implies that AVs might need to hide their identity to blend in visually and behaviorally as regular cars.

Blame Attribution Asymmetry in Human-Automation Cooperation

Human-automation cooperation has become ubiquitous. In this concept, automation refers to autonomous machines, robots, artificial intelligence, and other autonomous nonhuman agents. A human driver will share control of semiautonomous vehicles (semi-AVs) with an automated system and thus share responsibility for crashes caused by semi-AVs. Research has not clarified whether and why people would attribute different levels of blame and responsibility to automation (and its creators) and its human counterpart when each causes an equivalent crash. We conducted four experiments in two studies (total N = 1,045) to measure different responses (e.g., severity and acceptability judgment, blame and responsibility attribution, compensation judgment) to hypothetical crashes that are caused by the human or the automation in semi-AVs. The results provided previously unidentified evidence of a bias, which we called the "blame attribution asymmetry," a tendency that people will judge the automation-caused crash more harshly, ascribe more blame and responsibility to automation and its creators, and think the victim in this crash should be compensated more. This asymmetry arises in part because of the higher negative affect triggered by the automation-caused crash. This bias has a direct policy implication: a policy allowing "not-safe enough" semi-AVs on roads could backfire, because these AVs will lead to many traffic crashes, which might in turn produce greater psychological costs and deter more people from adopting them. Other theoretical and policy implications of our findings were also discussed.

With Clear Intention—An Ethical Responsibility Model for Robot Governance

There is much discussion about super artificial intelligence (AI) and autonomous machine learning (ML) systems, or learning machines (LM). Yet, the reality of thinking robotics still seems far on the horizon. It is one thing to define AI in light of human intelligence, citing the remoteness between ML and human intelligence, but another to understand issues of ethics, responsibility, and accountability in relation to the behavior of autonomous robotic systems within a human society. Due to the apparent gap between a society in which autonomous robots are a reality and present-day reality, many of the efforts placed on establishing robotic governance, and indeed, robot law fall outside the fields of valid scientific research. Work within this area has concentrated on manifestos, special interest groups and popular culture. This article takes a cognitive scientific perspective toward characterizing the nature of what true LMs would entail—i.e., intentionality and consciousness. It then proposes the Ethical Responsibility Model for Robot Governance (ER-RoboGov) as an initial platform or first iteration of a model for robot governance that takes the standpoint of LMs being conscious entities. The article utilizes past AI governance model research to map out the key factors of governance from the perspective of autonomous machine learning systems.

Comparative Analysis of Human Operators and Advanced Technologies in the Visual Inspection of Aero Engine Blades

Background—Aircraft inspection is crucial for safe flight operations and is predominantly performed by human operators, who are unreliable, inconsistent, subjective, and prone to err. Thus, advanced technologies offer the potential to overcome those limitations and improve inspection quality. Method—This paper compares the performance of human operators with image processing, artificial intelligence software and 3D scanning for different types of inspection. The results were statistically analysed in terms of inspection accuracy, consistency and time. Additionally, other factors relevant to operations were assessed using a SWOT and weighted factor analysis. Results—The results show that operators’ performance in screen-based inspection tasks was superior to inspection software due to their strong cognitive abilities, decision-making capabilities, versatility and adaptability to changing conditions. In part-based inspection however, 3D scanning outperformed the operator while being significantly slower. Overall, the strength of technological systems lies in their consistency, availability and unbiasedness. Conclusions—The performance of inspection software should improve to be reliably used in blade inspection. While 3D scanning showed the best results, it is not always technically feasible (e.g., in a borescope inspection) nor economically viable. This work provides a list of evaluation criteria beyond solely inspection performance that could be considered when comparing different inspection systems.

Blaming automated vehicles in difficult situations

Automated vehicles (AVs) have made huge strides toward large-scale deployment. Despite this progress, AVs continue to make mistakes, some resulting in death. Although some mistakes are avoidable, others are hard to avoid even by highly skilled drivers. As these mistakes continue to shape attitudes toward AVs, we need to understand whether people differentiate between them. We ask the following two questions. When an AV makes a mistake, does the perceived difficulty or novelty of the situation predict blame attributed to it? How does that blame attribution compare to a human driving a car? Through two studies, we find that the amount of blame people attribute to AVs and human drivers is sensitive to situation difficulty. However, while some situations could be more difficult for AVs and others for human drivers, people blamed AVs more, regardless. Our results provide novel insights in understanding psychological barriers influencing the public's view of AVs.

•

Attributed blame to machine and human drivers is sensitive to situation difficulty

•

Mistakes in simple situations receive more blame than in novel or complex situations

•

Machine drivers receive more blame, across different situations

Factors Influencing the Purchase Intention of Autonomous Cars

Even though autonomous cars have not yet crossed into the mainstream car market, their adoption seems inevitable, but not much is known about the purchasing intention of ACs and potential influences on it. To better understand the influences of various factors on purchasing intentions of autonomous cars, research using bibliometrics, an online survey and SEM modelling was performed. Based on an analysis of previous research work and the unified theory of acceptance of technology, an empirical model was produced and tested using data obtained from an online survey involving 266 individuals. The goal was to analyse which characteristics of autonomous cars, socio-demographic variables of potential buyers, and buyers’ personal and social characteristics could potentially influence the adoption of autonomous cars. The results show that factors of car safety, buyer age and level of education, perceived social influence, anxiety and performance expectancy are significantly correlated to purchasing intention of ACs, while correlations with other factors to purchasing intentions have not been proven.

How different autonomous vehicle presentation influences its acceptance: Is a communal car better than agentic one?

Public acceptance of autonomous vehicles (AVs) is still questionable. Nevertheless, it can be influenced by proper communication strategy. Therefore, our research focuses on (1) the type of information concerning AVs that consumers seek and (2) how to communicate this technology in order to increase its acceptance. In the first study (N = 711) topic modeling showed that the most sought for information concern the communion and the agency of AVs. In the second, experimental study (N = 303) we measured the participants' fear and goal-orientation in relation to AVs. Then, after the manipulation of the AV advertisement (imbued with communal vs agentic content), technology acceptance components (perceived ease of use, perceived usefulness and behavioral intention) were verified. The comparative analysis of the structural model estimates showed that the both participants' fear and goal-orientation in relation to AVs were associated much more with the acceptance components of the communal AV rather than the agentic one. Therefore, people want to know both whether AVs are communal and agentic, but they are more prone to accept a communal AV than agentic one.