Conference Agenda

Much has been made about the potential emergence of so-called ‘responsibility gaps’, either due to technology more generally or due to AI specifically. A multitude of responses have emerged in response to the responsibility gap challenge. An understudied response to the challenge, however, is vicarious responsibility as a potential bridge for responsibility gaps. In cases of vicarious responsibility, one agent stands as a substitute for another. The idea behind vicarious responsibility is therefore a simple one: it can helpfully account for cases where an agent is responsible for the uncontrollable and unpredictable actions of some other entity. Human agents, in certain circumstances, can stand in as responsible for the harms that follow from autonomous systems.

Surprisingly, however, very little attention has been given to the prospect of vicarious responsibility for autonomous AI-systems. With the exception of work by Trystan Goetze, who offers an argument in favour of vicarious responsibility for computing professionals in certain cases of AI-enabled harm, there is very little other literature on the intersection between digital ethics/AI ethics and theories of vicarious responsibility (2022).

Moreover, we find a similar lack of philosophical literature on vicarious responsibility more generally. While legal scholars debate its merit (usually in discussions of vicarious liability), there has only recently been spate of attention addressed to the philosophical underpinnings of vicarious responsibility (most of it in a special issue of the Monist) (Collins and De Haan, 2021; Goetze, 2021; Kuan, 2021; Mellor, 2021; Radoilska, 2021; Glavaničová and Pascucci, 2024). In this paper I aim to plug this gap and assess the limits and potential of vicarious responsibility as a solution to the responsibility gap challenge. To do so I proceed as follows.

First, I outline in general terms why the idea of vicarious responsibility makes sense, despite suggestions that it is a contradiction in terms.

Second, I outline two different ‘faces’ of responsibility – accountability and answerability, and show how responsibility as answerability can plausibly be borne vicariously. This is motivated by the idea that it seems unjustifiable to blame an agent for the deeds of another, but in certain cases it makes sense to expect one agent to answer for the actions of another.

Third, I describe why vicarious responsibility might be a useful analytical lens for understanding autonomous-systems. While Mellor (2021), for example, argues that vicarious responsibility can only exist between individuals (which I take to mean human persons), I think this overly restrictive. We can and should adopt a more holistic perspective on vicarious responsibility whereby such responsibility is fittingly attributable even in cases where only one entity involved is a human agent. An easy example of such a case is the responsibility I might have for my dog if he were to bite someone. In such a scenario, therefore, I am responsible for an unexpected and uncontrollable action that I did not initiate. I apply this discussion to autonomous systems and suggest that, under certain conditions, we are vicariously responsible, and thus answerable, for the actions of these entities.

Autonomous driving technology (ADT) is a key component of smart city development, gaining attention globally as governments experiment with its implementation through special zones. In Japan, ADT has been proposed as a solution to regional issues such as labor shortages and inadequate transportation due to an aging population and declining birthrate. Special zones provide a controlled environment for testing ADT and addressing regional challenges. However, this approach raises critical philosophical and institutional questions.

One major challenge lies in the lack of democratic transparency in the designation and governance of special zones. Questions such as who should make decisions, how diverse opinions are incorporated, and whether the process reflects public interest remain unresolved. Additionally, the inherent risk of failure in deploying ADT leads to ethical dilemmas: who bears responsibility for failure, and to what extent should society tolerate and learn from such failures?

Another critical issue is ensuring the long-term sustainability of ADT systems. While initial deployment often relies on government subsidies, the question arises as to whether autonomous transportation infrastructure can function under market principles once public support diminishes. Addressing this requires robust frameworks for public-private partnerships, cost-benefit analysis, and sustainable governance.

This paper examines ADT implementation in Japan’s special zones, such as Osaka and Kyushu, to explore the interplay between transparency, stakeholder engagement, and societal acceptance. By analyzing these cases, the study seeks to identify practical mechanisms to ensure democratic governance while managing the risks and uncertainties of ADT.

Finally, the paper engages with broader philosophical inquiries: How can societies build consensus around unknown technological innovations? To what extent should failure be accepted as a necessary condition for progress? These questions, while rooted in the specific context of ADT, resonate with larger concerns about the ethical and social implications of technology. Reflecting on these issues provides an opportunity to reconsider the role of technology in shaping future societies and the values that underpin its governance.