Robot Ethics 2.0

Robot Ethics 2.0 arrived from Oxford University Press in 2017 as a deliberate sequel. Five years earlier the same three editors — Patrick Lin, Keith Abney and Ryan Jenkins, all based at California Polytechnic State University’s Ethics + Emerging Sciences Group — had assembled Robot Ethics: The Ethical and Social Implications of Robotics, a volume widely credited with helping establish the field as something philosophers, engineers and policy writers could cite in the same bibliography. By 2017 the ground had shifted. Google’s self-driving project had become Waymo. Tesla’s Autopilot had killed its first driver. The United States was flying armed drones in six countries. DARPA had run its Robotics Challenge. The Campaign to Stop Killer Robots was into its fourth year. The second volume collects more than two dozen essays from philosophers, legal scholars, roboticists and ethicists to catch the discussion up to this new reality, and to push into questions the first book had left open.

The editors do not advance a single thesis, as is the way of an edited collection, but they do set a common frame. Robots and artificial agents are leaving the controlled settings of the factory floor and the research lab and entering domains — public roads, hospitals, battlefields, bedrooms, nursing homes — where their decisions touch human beings in morally loaded ways. The book’s working premise is that this transition cannot be handled by the engineering disciplines alone. Writing software that behaves well on a quiet motorway is not the same as writing software that chooses whom to swerve toward when a crash is unavoidable. Specifying a weapon that strikes a target is not the same as specifying one that selects the target. The contributors argue, in different registers and with different conclusions, that these are irreducibly ethical questions and that leaving them to accrete through litigation and precedent is a choice with consequences.

The volume moves in sections. An opening cluster takes up moral and legal theory as it applies to artificial agents in general — the problem of whether a robot can be a moral patient, the problem of responsibility gaps when harm is caused by a system no single human directed, the problem of programming values into a machine when ethicists themselves disagree about the values. A substantial section is given to autonomous vehicles, which in 2017 were the field’s most immediate test case: a technology sold to the public as imminent, with regulators in several US states already writing rules for it. The military chapters sit at the centre of the book. A further cluster takes up medical and care robotics, including surgical systems and the robots being piloted in Japanese and European elder-care facilities. There are chapters on sex robots, on employment and economic displacement, and on the governance and liability questions that cut across every application.

The autonomous-vehicle chapters are where the book is at its most accessible, and where it did some of its most quoted work. Several contributors engage with the trolley-problem analogy — the thought experiment, originally from Philippa Foot, in which a runaway tram will kill five people unless diverted to a side track where it will kill one. Transposing this to a self-driving car produces vivid variants: the car must choose between its own occupant and a pedestrian, between a child and an adult, between a motorcyclist with a helmet and one without. Jason Millar’s contribution on what he calls “proxy agency” argues that these choices should not be the designer’s to make silently, and that drivers should retain some say in the ethical settings of a vehicle that acts in their name. Other authors push back, noting that trolley cases are a poor guide to the statistical business of actually reducing road deaths — a car that chooses well in a thousand everyday near-misses matters more than one that handles a philosopher’s dilemma well. The disagreement is representative of how the book treats most topics: competing arguments presented in enough depth that a reader can see the shape of the live dispute.

The military chapters are the ones most closely watched outside the academy. Ronald Arkin, the Georgia Tech roboticist whose earlier work with the US Army on an “ethical governor” for autonomous weapons had drawn both interest and alarm, contributes an argument for conditional acceptance: machines, Arkin maintains, can be built to adhere more strictly to the laws of armed conflict than human soldiers under stress, and the right response is to insist on that engineering rather than to foreclose the category. Against him, Peter Asaro and Noel Sharkey — both veterans of the International Committee for Robot Arms Control — argue that delegating the kill decision to software crosses a line that international humanitarian law has always kept with the human. Their chapters set out the familiar objections: machines cannot read context the way a soldier can, cannot be held accountable in the way a soldier can, cannot refuse an unlawful order, and in large numbers would lower the political cost of starting wars. Related chapters examine how just war theory — jus ad bellum and jus in bello — handles weapons that select and engage targets without a human in the firing loop, and how the Martens Clause and the principle of distinction constrain what algorithmic targeting may do. The treatment is sober rather than partisan: both the “ban” position associated with the Campaign to Stop Killer Robots and the “regulate” position associated with militaries and defence contractors are given room to make their best case.

Several of these chapters anchor their arguments in named systems rather than abstractions. The South Korean SGR-A1 sentry robot deployed along the DMZ, Israel’s Harpy and Harop loitering munitions which strike radar emitters without a human-in-the-loop at the moment of engagement, the US Navy’s Phalanx close-in weapon system which has had an autonomous mode since the 1980s, and the X-47B carrier-launched demonstrator all appear in the discussion. So does the US Department of Defense Directive 3000.09, the 2012 instruction that first set internal policy on autonomy in weapon systems. The chapters read, in retrospect, as the moment the debate shifted from hypothetical to inventory-based — from “if we build these” to “which existing systems already cross which lines.”

Away from the battlefield, the care-robot and medical-robot chapters do some of the volume’s most interesting work on the question of relational harm. Shannon Vallor’s contribution on robots in elder care argues that even a well-functioning machine companion may erode practices of attention and solidarity that give caregiving its moral weight; the chapter sits in a longer tradition of worry about technology substituting for presence. Other contributors are less pessimistic, pointing to pilot studies with PARO, the therapeutic seal robot developed at Japan’s AIST, as evidence that machines can meaningfully reduce agitation and loneliness in dementia patients where human staff are scarce. Surgical systems like Intuitive Surgical’s da Vinci are discussed as a case where the ethical questions turn less on autonomy — the surgeon still operates — than on skill transfer, training pipelines and the opacity of a device whose maker holds the service contract.

The sex-robot chapter, by John Danaher, became one of the book’s most cited pieces and generated its own secondary literature. Danaher is cautious of both the enthusiasts who see in sex robots a solution to loneliness and isolation and the critics who see in them an entrenchment of pornographic scripts about women; the chapter catalogues the arguments on each side and declines to settle them, which is the editorial posture of the collection more generally. Related chapters on friendship with artificial agents and on the moral status of entities that merely appear sentient join a longer-running debate between those who, like David Levy, see no bright line and those who, like Sherry Turkle, see one being eroded for no good reason.

The book’s section on machine ethics proper — on whether and how to build moral reasoning into software — surveys the main approaches without endorsing one. Top-down implementations try to encode a specific framework (Kantian deontology, utilitarian calculus, Asimov-style rules) and derive behaviour from it; bottom-up implementations use machine learning to extract norms from examples of human behaviour; hybrid architectures try to combine the two. Wendell Wallach and Colin Allen’s earlier Moral Machines (2008) framework hovers over the discussion. The limits are acknowledged honestly: no approach handles novel moral situations as well as a competent human; most approaches degrade in the presence of adversarial inputs; and benchmarks for moral performance barely exist. The chapters anticipate questions that would become urgent two or three years later, when large language models made the gap between fluent moral talk and reliable moral behaviour impossible to ignore.

Reception within the field has been broadly favourable, with the collection widely assigned on graduate syllabi in philosophy of technology, applied ethics and emerging weapons law. The book is treated as a reference point rather than a last word. Scholars sympathetic to a preemptive ban on lethal autonomous weapons have cited Asaro and Sharkey’s chapters in the Group of Governmental Experts discussions at the UN Convention on Certain Conventional Weapons; those more comfortable with regulation have returned to Arkin to argue the technical case for compliance by design. The autonomous-vehicle chapters have been engaged with by regulators at the US National Highway Traffic Safety Administration and in European Union documents on automated driving ethics. Some of the sharpest pushback has come from scholars arguing that the trolley-problem framing crowded out more important questions about data governance, surveillance and the political economy of mobility — a critique the editors have, in subsequent interviews, treated as fair.

For a reader working through the literature on AI in war today, Robot Ethics 2.0 is best read as the ethics-of-robotics complement to the political and operational books that followed it. Paul Scharre’s Army of None, published a year later, covers overlapping ground from the inside of the US defence policy conversation and will give a reader more by way of named programmes and procurement politics. Peter Singer’s earlier Wired for War set the journalistic scene. Armin Krishnan’s Killer Robots sets the legal frame. What this book adds, and what those don’t, is the philosophical machinery — the categories a reader needs to distinguish a problem of autonomy from a problem of agency, a problem of responsibility from a problem of liability, a problem of distinction from a problem of proportionality. It is not a book about Russia’s Lancet-3 loitering munition or China’s CH-5 or the Shahed-136, none of which were household names at publication. It is a book about the concepts a reader needs to think clearly about any such system.

Some parts of the book are already of their time. The autonomous-vehicle chapters assume a timeline for mass deployment that has slipped by close to a decade. The machine-ethics discussion predates the generative-AI boom and treats natural-language reasoning as a far distant prospect; several of its examples read differently now. The sex-robot material is anchored in a market that remains smaller than the chapter anticipated. What has aged better is the structural analysis — the distinction between designer, deployer and user responsibility; the taxonomy of autonomy from supervised to full; the catalogue of ways a well-engineered machine can still cause morally significant harm. These frames continue to show up, often unattributed, in the policy documents and ethics guidelines that governments and companies now publish on an almost monthly basis, which is the quiet form of influence most academic volumes aspire to.