Human in the Loop

In 2018 millions of people worldwide were playing a disturbing game. On their screens, a self-driving car with failed brakes hurtles towards an unavoidable collision. The choice is stark: plough straight ahead and kill three elderly pedestrians crossing legally, or swerve into a concrete barricade and kill three young passengers buckled safely inside. Click left. Click right. Save the young. Save the old. Each decision takes seconds, but the implications stretch across philosophy, engineering, law, and culture. The game was called the Moral Machine, and whilst it may have looked like entertainment, it's actually the largest global ethics experiment ever conducted. Designed by researchers Edmond Awad, Iyad Rahwan, and their colleagues at the Massachusetts Institute of Technology's Media Lab, it was built to answer a question that's become urgently relevant as autonomous vehicles edge closer to our roads: when AI systems make life-and-death decisions, whose moral values should they reflect?

The results, published in Nature in October 2018, were as fascinating as they were troubling. Over 40 million decisions from 233 countries and territories revealed not a unified human morality, but a fractured ethical landscape where culture, economics, and geography dramatically shape our moral intuitions. In some countries, participants overwhelmingly chose to spare the young over the elderly. In others, the preference was far less pronounced. Some cultures prioritised pedestrians; others favoured passengers. The study, conducted by Edmond Awad, Iyad Rahwan, and colleagues, exposed an uncomfortable truth: there is no universal answer to the trolley problem when it's rolling down real streets in the form of a two-tonne autonomous vehicle.

This isn't merely an academic exercise. Waymo operates robotaxi services in several American cities. Tesla's “Full Self-Driving” system (despite its misleading name) navigates city streets. Chinese tech companies are racing ahead with autonomous bus trials. The technology is here, imperfect and improving, and it needs ethical guidelines. The question is no longer whether autonomous vehicles will face moral dilemmas, but who gets to decide how they're resolved.

The Trolley Problem

The classic trolley problem, formulated by philosopher Philippa Foot in 1967, was never meant to be practical. It was a thought experiment, a tool for probing the boundaries between utilitarian and deontological ethics. But autonomous vehicles have dragged it kicking and screaming into the real world, where abstract philosophy collides with engineering specifications, legal liability, and consumer expectations.

The Moral Machine experiment presented participants with variations of the scenario in which an autonomous vehicle's brakes have failed. Thirteen factors were tested across different combinations: should the car spare humans over pets, passengers over pedestrians, more lives over fewer, women over men, the young over the elderly, the fit over the infirm, those of higher social status over lower, law-abiders over law-breakers? And crucially: should the car swerve (take action) or stay its course (inaction)?

The global preferences revealed by the data showed some universal trends. Across nearly all cultures, participants preferred sparing humans over animals and sparing more lives over fewer. But beyond these basics, consensus evaporated. The study identified three major cultural clusters with distinct ethical preferences: Western countries (including North America and many European nations), Eastern countries (including many Asian nations grouped under the problematic label of “Confucian” societies), and Southern countries (including Latin America and some countries with French influence).

These weren't minor differences. Participants from collectivist cultures like China and Japan showed far less preference for sparing the young over the elderly compared to individualistic Western cultures. The researchers hypothesised this reflected cultural values around respecting elders and the role of the individual versus the community. Meanwhile, participants from countries with weaker rule of law were more tolerant of jaywalkers versus pedestrians crossing legally, suggesting that lived experience with institutional strength shapes ethical intuitions.

Economic inequality also left its fingerprints on moral choices. Countries with higher levels of economic inequality showed greater gaps in how they valued individuals of high versus low social status. It's a sobering finding: the moral values we encode into machines may reflect not our highest ideals, but our existing social prejudices.

The scale of the Moral Machine experiment itself tells a story about global interest in these questions. When the platform launched in 2014, the researchers at MIT expected modest participation. Instead, it went viral across social media, translated into ten languages, and became a focal point for discussions about AI ethics worldwide. The 40 million decisions collected represent the largest dataset ever assembled on moral preferences across cultures. Participants weren't just clicking through scenarios; many spent considerable time deliberating, revisiting choices, and engaging with the ethical complexity of each decision.

Yet for all its scope, the Moral Machine has limitations that its creators readily acknowledge. The scenarios present artificial constraints that rarely occur in reality. The experiment assumes autonomous vehicles will face genuine no-win situations where harm is unavoidable. In practice, advanced AI systems should be designed to avoid such scenarios entirely through superior sensing, prediction, and control. The real question may not be “who should the car kill?” but rather “how can we design systems that never face such choices?”

However, the trolley problem may turn out to be the least important problem of all.

The Manufacturer's Dilemma

For automotive manufacturers, the Moral Machine results present a nightmare scenario. Imagine you're an engineer at Volkswagen's autonomous vehicle division in Germany. You're programming the ethical decision-making algorithm for a car that will be sold globally. Do you optimise it for German preferences? Chinese preferences? American preferences? A global average that satisfies no one?

The engineering challenge is compounded by a fundamental mismatch between how the trolley problem is framed and how autonomous vehicles actually operate. The Moral Machine scenarios assume perfect information: the car knows exactly how many people are in each group, their ages, whether they're obeying traffic laws. Real-world computer vision systems don't work that way. They deal in probabilities and uncertainties. A pedestrian detection system might be 95 per cent confident that object is a human, 70 per cent confident about their approximate age range, and have no reliable way to assess their social status or physical fitness.

Moreover, the scenarios assume binary choices and unavoidable collisions. Real autonomous vehicles operate in a continuous decision space, constantly adjusting speed, position, and trajectory to maximise safety for everyone. The goal isn't to choose who dies, it's to create a probability distribution of outcomes that minimises harm across all possibilities. As several robotics researchers have pointed out, the trolley problem may be asking the wrong question entirely.

Yet manufacturers can't simply ignore the ethical dimensions. Every decision about how an autonomous vehicle's software weights different factors, how it responds to uncertainty, how it balances passenger safety versus pedestrian safety, embeds ethical values. Those values come from somewhere. Currently, they largely come from the engineering teams and the corporate cultures within which they work.

In 2016, Mercedes-Benz caused controversy when a company executive suggested their autonomous vehicles would prioritise passenger safety over pedestrians in unavoidable collision scenarios. The company quickly clarified its position, but the episode revealed the stakes. If manufacturers openly prioritise their customers' safety over others, it could trigger a race to the bottom, with each company trying to offer the most “protective” system. The result might be vehicles that collectively increase risk for everyone outside a car whilst competing for the loyalty of those inside.

Some manufacturers have sought external guidance. In 2017, Germany's Federal Ministry of Transport and Digital Infrastructure convened an ethics commission to develop guidelines for automated and connected driving. The commission's report emphasised that human life always takes priority over property and animal life, and that distinctions based on personal features such as age, gender, or physical condition are strictly prohibited. It was an attempt to draw clear lines, but even these principles leave enormous room for interpretation when translated into code.

The German guidelines represent one of the most thorough governmental attempts to grapple with autonomous vehicle ethics. The 20 principles cover everything from data protection to the relationship between human and machine decision-making. Guideline 9 states explicitly: “In hazardous situations that prove to be unavoidable, the protection of human life enjoys top priority in a balancing of legally protected interests. Thus, within the constraints of what is technologically feasible, the objective must be to avoid personal injury.” It sounds clear, but the phrase “within the constraints of what is technologically feasible” opens significant interpretive space.

The commission also addressed accountability, stating that while automated systems can be tools to help people, responsibility for decisions made by the technology remains with human actors. This principle, whilst philosophically sound, creates practical challenges for liability frameworks. When an autonomous vehicle operating in fully automated mode causes harm, tracing responsibility back through layers of software, hardware, training data, and corporate decision-making becomes extraordinarily complex.

Meanwhile, manufacturers are making these choices in relative silence. The algorithms governing autonomous vehicle behaviour are proprietary, protected as trade secrets. We don't know precisely how Tesla's system prioritises different potential outcomes, or how Waymo's vehicles weight passenger safety against pedestrian safety. This opacity makes democratic oversight nearly impossible and prevents meaningful public debate about the values embedded in these systems.

The Owner's Perspective

What if the car's owner got to choose? It's an idea that has appeal on the surface. After all, you own the vehicle. You're legally responsible for it in most jurisdictions. Shouldn't you have a say in its ethical parameters?

This is where things get truly uncomfortable. Research conducted at the University of California, Berkeley, and elsewhere has shown that people's ethical preferences change dramatically depending on whether they're asked about “cars in general” or “my car.” When asked about autonomous vehicles as a societal technology, people tend to endorse utilitarian principles: save the most lives, even if it means sacrificing the passenger. But when asked what they'd want from a car they'd actually purchase for themselves and their family, preferences shift sharply towards self-protection.

It's a version of the classic collective action problem. Everyone agrees that in general, autonomous vehicles should minimise total casualties. But each individual would prefer their specific vehicle prioritise their survival. If manufacturers offered this as a feature, they'd face a catastrophic tragedy of the commons. Roads filled with self-protective vehicles would be less safe for everyone.

There's also the thorny question of what “personalised ethics” would even mean in practice. Would you tick boxes in a configuration menu? “In unavoidable collision scenarios, prioritise: (a) occupants, (b) minimise total casualties, © protect children”? It's absurd on its face, yet the alternative, accepting whatever ethical framework the manufacturer chooses, feels uncomfortably like moral outsourcing.

The legal implications are staggering. If an owner has explicitly configured their vehicle to prioritise their safety over pedestrians, and the vehicle then strikes and kills a pedestrian in a scenario where a different setting might have saved them, who bears responsibility? The owner, for their configuration choice? The manufacturer, for offering such choices? The software engineers who implemented the feature? These aren't hypothetical questions. They're exactly the kind of liability puzzles that will land in courts within the next decade.

Some researchers have proposed compromise positions: allow owners to choose between a small set of ethically vetted frameworks, each certified as meeting minimum societal standards. But this just pushes the question back a level: who decides what's ethically acceptable? Who certifies the certifiers?

The psychological dimension of ownership adds further complexity. Studies in behavioural economics have shown that people exhibit strong “endowment effects,” valuing things they own more highly than identical things they don't own. Applied to autonomous vehicles, this suggests owners might irrationally overvalue the safety of their vehicle's occupants compared to others on the road. It's not necessarily conscious bias; it's a deep-seated cognitive tendency that affects how we weigh risks and benefits.

There's also the question of what happens when ownership itself becomes murky. Autonomous vehicles may accelerate the shift from ownership to subscription and shared mobility services. If you don't own the car but simply summon it when needed, whose preferences should guide its ethical parameters? The service provider's? An aggregate of all users? Your personal profile built from past usage? The more complex ownership and usage patterns become, the harder it is to assign moral authority over the vehicle's decision-making.

Insurance companies, too, have a stake in these questions. Actuarial calculations for autonomous vehicles will need to account for the ethical frameworks built into their software. A vehicle programmed with strong passenger protection might command higher premiums for third-party liability coverage. These economic signals could influence manufacturer choices in ways that have nothing to do with philosophical ethics and everything to do with market dynamics.

Society's Stake

If the decision can't rest with manufacturers (too much corporate interest) or owners (too much self-interest), perhaps it should be made by society collectively through democratic processes. This is the argument advanced by many ethicists and policy researchers. Autonomous vehicles operate in shared public space. Their decisions affect not just their occupants but everyone around them. That makes their ethical parameters a matter for collective deliberation and democratic choice.

In theory, it's compelling. In practice, it's fiendishly complicated. Start with the question of jurisdiction. Traffic laws are national, but often implemented at state or local levels, particularly in federal systems like the United States, Germany, or Australia. Should ethical guidelines for autonomous vehicles be set globally, nationally, regionally, or locally? The Moral Machine data suggests that even within countries, there can be significant ethical diversity.

Then there's the challenge of actually conducting the deliberation. Representative democracy works through elected officials, but the technical complexity of autonomous vehicle systems means that most legislators lack the expertise to meaningfully engage with the details. Do you defer to expert committees? Then you're back to a technocratic solution that may not reflect public values. Do you use direct democracy, referendums on specific ethical parameters? That's how Switzerland handles many policy questions, but it's slow, expensive, and may not scale to the detailed, evolving decisions needed for AI systems.

Several jurisdictions have experimented with middle paths. The German ethics commission mentioned earlier included philosophers, lawyers, engineers, and civil society representatives. Its 20 guidelines attempted to translate societal values into actionable principles for autonomous driving. Among them: automated systems must not discriminate on the basis of individual characteristics, and in unavoidable accident scenarios, any distinction based on personal features is strictly prohibited.

But even this well-intentioned effort ran into problems. The prohibition on discrimination sounds straightforward, but autonomous vehicles must make rapid decisions based on observable characteristics. Is it discriminatory for a car to treat a large object differently from a small one? That distinction correlates with age. Is it discriminatory to respond differently to an object moving at walking speed versus running speed? That correlates with fitness. The ethics become entangled with the engineering in ways that simple principles can't cleanly resolve.

There's also a temporal problem. Democratic processes are relatively slow. Technology evolves rapidly. By the time a society has deliberated and reached consensus on ethical guidelines for current autonomous vehicle systems, the technology may have moved on, creating new ethical dilemmas that weren't anticipated. Some scholars have proposed adaptive governance frameworks that allow for iterative refinement, but these require institutional capacity that many jurisdictions simply lack.

Public deliberation efforts that have been attempted reveal the challenges. In 2016, researchers at the University of California, Berkeley conducted workshops where citizens were presented with autonomous vehicle scenarios and asked to deliberate on appropriate responses. Participants struggled with the technical complexity, often reverting to simplified heuristics that didn't capture the nuances of real-world scenarios. When presented with probabilistic information (the system is 80 per cent certain this object is a child), many participants found it difficult to formulate clear preferences.

The challenge of democratic input is compounded by the problem of time scales. Autonomous vehicle technology is developing over years and decades, but democratic attention is sporadic and driven by events. A high-profile crash involving an autonomous vehicle might suddenly focus public attention and demand immediate regulatory response, potentially leading to rules formed in the heat of moral panic rather than careful deliberation. Conversely, in the absence of dramatic incidents, the public may pay little attention whilst crucial decisions are made by default.

Some jurisdictions are experimenting with novel forms of engagement. Citizens' assemblies, where randomly selected members of the public are brought together for intensive deliberation on specific issues, have been used in Ireland and elsewhere for contentious policy questions. Could similar approaches work for autonomous vehicle ethics? The model has promise, but scaling it to address the range of decisions needed across different jurisdictions presents formidable challenges.

No Universal Morality

Perhaps the most unsettling implication of the Moral Machine study is that there may be no satisfactory global solution. The ethical preferences revealed by the data aren't merely individual quirks; they're deep cultural patterns rooted in history, religion, economic development, and social structure.

The researchers found that countries clustered into three broad groups based on their moral preferences. The Western cluster, including the United States, Canada, and much of Europe, showed strong preferences for sparing the young over the elderly, for sparing more lives over fewer, and generally exhibited what the researchers characterised as more utilitarian and individualistic patterns. The Eastern cluster, including Japan and several other Asian countries, showed less pronounced preferences for sparing the young and patterns suggesting more collectivist values. The Southern cluster, including many Latin American and some Middle Eastern countries, showed distinct patterns again.

These aren't value judgements about which approach is “better.” They're empirical observations about diversity. But they create practical problems for a globalised automotive industry. A car engineered according to Western ethical principles might behave in ways that feel wrong to drivers in Eastern countries, and vice versa. The alternative, creating region-specific ethical programming, raises uncomfortable questions about whether machines should be designed to perpetuate cultural differences in how we value human life.

There's also the risk of encoding harmful biases. The Moral Machine study found that participants from countries with higher economic inequality showed greater willingness to distinguish between individuals of high and low social status when making life-and-death decisions. Should autonomous vehicles in those countries be programmed to reflect those preferences? Most ethicists would argue absolutely not, that some moral principles (like the equal value of all human lives) should be universal regardless of local preferences.

But that introduces a new problem: whose ethics get to be universal? The declaration that certain principles override cultural preferences is itself a culturally situated claim, one that has historically been used to justify various forms of imperialism and cultural dominance. The authors of the Moral Machine study were careful to note that their results should not be used to simply implement majority preferences, particularly where those preferences might violate fundamental human rights or dignity.

The geographic clustering in the data reveals patterns that align with existing cultural frameworks. Political scientists Ronald Inglehart and Christian Welzel's “cultural map of the world” divides societies along dimensions of traditional versus secular-rational values and survival versus self-expression values. When the Moral Machine data was analysed against this framework, strong correlations emerged. Countries in the “Protestant Europe” cluster showed different patterns from those in the “Confucian” cluster, which differed again from the “Latin America” cluster.

These patterns aren't random. They reflect centuries of historical development, religious influence, economic systems, and political institutions. The question is whether autonomous vehicles should perpetuate these differences or work against them. If Japanese autonomous vehicles are programmed to show less preference for youth over age, reflecting Japanese cultural values around elder respect, is that celebrating cultural diversity or encoding ageism into machines?

The researchers themselves wrestled with this tension. In their Nature paper, Awad, Rahwan, and colleagues wrote: “We do not think that the preferences revealed in the Moral Machine experiment should be directly translated into algorithmic rules... Cultural preferences might not reflect what is ethically acceptable.” It's a crucial caveat that prevents the study from becoming a simple guide to programming autonomous vehicles, but it also highlights the gap between describing moral preferences and prescribing ethical frameworks.

Beyond the Trolley

Focusing on trolley-problem scenarios may actually distract from more pressing and pervasive ethical issues in autonomous vehicle development. These aren't about split-second life-and-death dilemmas but about the everyday choices embedded in the technology.

Consider data privacy. Autonomous vehicles are surveillance systems on wheels, equipped with cameras, lidar, radar, and other sensors that constantly monitor their surroundings. This data is potentially valuable for improving the systems, but it also raises profound privacy concerns. Who owns the data about where you go, when, and with whom? How long is it retained? Who can access it? These are ethical questions, but they're rarely framed that way.

Or consider accessibility and equity. If autonomous vehicles succeed in making transportation safer and more efficient, but they remain expensive luxury goods, they could exacerbate existing inequalities. Wealthy neighbourhoods might become safer as autonomous vehicles replace human drivers, whilst poorer areas continue to face higher traffic risks. The technology could entrench a two-tier system where your access to safe transportation depends on your income.

Then there's the question of employment. Driving is one of the most common occupations in many countries. Millions of people worldwide earn their living as taxi drivers, lorry drivers, delivery drivers. The widespread deployment of autonomous vehicles threatens this employment, with cascading effects on families and communities. The ethical question isn't just about building the technology, but about managing its social impact.

Environmental concerns add another layer. Autonomous vehicles could reduce emissions if they're electric and efficiently managed through smart routing. Or they could increase total vehicle miles travelled if they make driving so convenient that people abandon public transport. The ethical choices about how to deploy and regulate the technology will have climate implications that dwarf the trolley problem.

The employment impacts deserve deeper examination. In the United States alone, approximately 3.5 million people work as truck drivers, with millions more employed as taxi drivers, delivery drivers, and in related occupations. Globally, the numbers are far higher. The transition to autonomous vehicles won't happen overnight, but when it does accelerate, the displacement could be massive and concentrated in communities that already face economic challenges.

This isn't just about job losses; it's about the destruction of entire career pathways. Driving has traditionally been one avenue for people without advanced education to earn middle-class incomes. If that pathway closes without adequate alternatives, the social consequences could be severe. Some economists argue that new jobs will emerge to replace those lost, as has happened with previous waves of automation. But the timing, location, and skill requirements of those new jobs may not align with the needs of displaced workers.

The ethical responsibility for managing this transition doesn't rest solely with autonomous vehicle manufacturers. It's a societal challenge requiring coordinated policy responses: education and retraining programmes, social safety nets, economic development initiatives for affected communities. But the companies developing and deploying the technology bear some responsibility for the consequences of their innovations. How much? That's another contested ethical question.

Data privacy concerns aren't merely about consumer protection; they involve questions of power and control. Autonomous vehicles will generate enormous amounts of data about human behaviour, movement patterns, and preferences. This data has tremendous commercial value for targeted advertising, urban planning, real estate development, and countless other applications. Who owns this data? Who profits from it? Who gets to decide how it's used?

Current legal frameworks around data ownership are ill-equipped to handle the complexities. In some jurisdictions, data generated by a device belongs to the device owner. In others, it belongs to the service provider or manufacturer. The European Union's General Data Protection Regulation provides some protections, but many questions remain unresolved. When your autonomous vehicle's sensors capture images of pedestrians, who owns that data? The pedestrians certainly didn't consent to being surveilled.

There's also the problem of data security. Autonomous vehicles are computers on wheels, vulnerable to hacking like any networked system. A compromised autonomous vehicle could be weaponised, used for surveillance, or simply disabled. The ethical imperative to secure these systems against malicious actors is clear, but achieving robust security whilst maintaining the connectivity needed for functionality presents ongoing challenges.

These broader ethical challenges, whilst less dramatic than the trolley problem, are more immediate and pervasive. They affect every autonomous vehicle on every journey, not just in rare emergency scenarios. The regulatory frameworks being developed need to address both the theatrical moral dilemmas and the mundane but consequential ethical choices embedded throughout the technology's deployment.

Regulation in the Real World

Several jurisdictions have begun grappling with these issues through regulation, with varying approaches. In the United States, the patchwork of state-level regulations has created a complex landscape. California, Arizona, and Nevada have been particularly active in welcoming autonomous vehicle testing, whilst other states have been more cautious. The federal government has issued guidance but largely left regulation to states.

The European Union has taken a more coordinated approach, with proposals for continent-wide standards that would ensure autonomous vehicles meet common safety and ethical requirements. The aforementioned German ethics commission's guidelines represent one influential model, though their translation into binding law remains incomplete.

China, meanwhile, has pursued rapid development with significant state involvement. Chinese companies and cities have launched ambitious autonomous vehicle trials, but the ethical frameworks guiding these deployments are less transparent to outside observers. The country's different cultural values around privacy, state authority, and individual rights create a distinct regulatory environment.

What's striking about these early regulatory efforts is how much they've focused on technical safety standards (can the vehicle detect obstacles? Does it obey traffic laws?) and how little on the deeper ethical questions. This isn't necessarily a failure; it may reflect a pragmatic recognition that we need to solve basic safety before tackling philosophical dilemmas. But it also means we're building infrastructure and establishing norms without fully addressing the value questions at the technology's core.

The regulatory divergence between jurisdictions creates additional complications for manufacturers operating globally. An autonomous vehicle certified for use in California may not meet German standards, which differ from Chinese requirements. These aren't just technical specifications; they reflect different societal values about acceptable risk, privacy, and the relationship between state authority and individual autonomy.

Some industry advocates have called for international harmonisation of autonomous vehicle standards, similar to existing frameworks for aviation. The International Organisation for Standardisation and the United Nations Economic Commission for Europe have both initiated efforts in this direction. But harmonising technical standards is far easier than harmonising ethical frameworks. Should the international standard reflect Western liberal values, Confucian principles, Islamic ethics, or some attempted synthesis? The very question reveals the challenge.

Consider testing and validation. Before an autonomous vehicle can be deployed on public roads, regulators need assurance that it meets safety standards. But how do you test for ethical decision-making? You can simulate scenarios, but the Moral Machine experiment demonstrated that people disagree about the “correct” answers. If a vehicle consistently chooses to protect passengers over pedestrians, is that a bug or a feature? The answer depends on your ethical framework.

Some jurisdictions have taken the position that autonomous vehicles should simply be held to the same standards as human drivers. If they cause fewer crashes and fatalities than human-driven vehicles, they've passed the test. This approach sidesteps the trolley problem by focusing on aggregate outcomes rather than individual ethical decisions. It's pragmatic, but it may miss important ethical dimensions. A vehicle that reduces total harm but does so through systemic discrimination might be statistically safer but ethically problematic.

Transparency and Ongoing Deliberation

If there's no perfect answer to whose morals should guide autonomous vehicles, perhaps the best approach is radical transparency combined with ongoing public deliberation. Instead of trying to secretly embed a single “correct” ethical framework, manufacturers and regulators could make their choices explicit and subject to democratic scrutiny.

This would mean publishing the ethical principles behind autonomous vehicle decision-making in clear, accessible language. It would mean creating mechanisms for public input and regular review. It would mean acknowledging that these are value choices, not purely technical ones, and treating them accordingly.

Some progress is being made in this direction. The IEEE, a major professional organisation for engineers, has established standards efforts around ethical AI development. Academic institutions are developing courses in technology ethics that integrate philosophical training with engineering practice. Some companies have created ethics boards to review their AI systems, though the effectiveness of these bodies varies widely.

What's needed is a culture shift in how we think about deploying AI systems in high-stakes contexts. The default mode in technology development has been “move fast and break things,” with ethical considerations treated as afterthoughts. For autonomous vehicles, that approach is inadequate. We need to move deliberately, with ethical analysis integrated from the beginning.

This doesn't mean waiting for perfect answers before proceeding. It means being honest about uncertainty, building in safeguards, and creating robust mechanisms for learning and adaptation. It means recognising that the question of whose morals should guide autonomous vehicles isn't one we'll answer once and for all, but one we'll need to continually revisit as the technology evolves and as our societal values develop.

The Moral Machine experiment demonstrated that human moral intuitions are diverse, context-dependent, and shaped by culture and experience. Rather than seeing this as a problem to be solved, we might recognise it as a feature of human moral reasoning. The challenge isn't to identify the single correct ethical framework and encode it into our machines. The challenge is to create systems, institutions, and processes that can navigate this moral diversity whilst upholding fundamental principles of human dignity and rights.

Autonomous vehicles are coming. The technology will arrive before we've reached consensus on all the ethical questions it raises. That's not an excuse for inaction, but a call for humility, transparency, and sustained engagement. The cars will drive themselves, but the choice of whose values guide them? That remains, must remain, a human decision. And it's one we'll be making and remaking for years to come.

One thing is certain, however. The ethics of autonomous vehicles may be like the quest for a truly random number: something we can approach, simulate, and refine, but never achieve in the pure sense. Some questions are not meant to be answered, only continually debated.

---

Sources and References

1. Awad, E., Dsouza, S., Kim, R., Schulz, J., Henrich, J., Shariff, A., Bonnefon, J.-F., & Rahwan, I. (2018). The Moral Machine experiment. Nature, 563, 59–64. https://doi.org/10.1038/s41586-018-0637-6

2. MIT Technology Review. (2018, October 24). Should a self-driving car kill the baby or the grandma? Depends on where you're from. https://www.technologyreview.com/2018/10/24/139313/a-global-ethics-study-aims-to-help-ai-solve-the-self-driving-trolley-problem/

3. Bonnefon, J.-F., Shariff, A., & Rahwan, I. (2016). The social dilemma of autonomous vehicles. Science, 352(6293), 1573–1576. https://doi.org/10.1126/science.aaf2654

4. Federal Ministry of Transport and Digital Infrastructure, Germany. (2017). Ethics Commission: Automated and Connected Driving. Report presented in Berlin, June 2017.

---

Tim Green UK-based Systems Theorist & Independent Technology Writer

Tim explores the intersections of artificial intelligence, decentralised cognition, and posthuman ethics. His work, published at smarterarticles.co.uk, challenges dominant narratives of technological progress while proposing interdisciplinary frameworks for collective intelligence and digital stewardship.

His writing has been featured on Ground News and shared by independent researchers across both academic and technological communities.

ORCID: 0009-0002-0156-9795 Email: tim@smarterarticles.co.uk

Discuss...