Responsible Innovation in the Digital World: Designing AI with Moral Values

Short Summary

In this Royal Institution talk, the speaker argues that responsible innovation in AI and digital technologies must be designed by values. Tracing the term's origin in The Netherlands, he outlines Europe’s human-centered approach to AI and warns against pursuing progress without ethical safeguards. The talk presents a practical framework that translates high level values into concrete design requirements, with compelling Dutch examples such as modular, privacy preserving devices and innovative urban designs. The central message is that ethics must be embedded in the design process from the start to maximize benefits and minimize harms, guiding technology toward solutions that truly serve humanity.

Introduction and the Core Idea: Responsible Innovation Through Design

The talk opens by connecting responsible innovation in the digital world to its Dutch roots, explaining that the concept was invented in The Hague by the Dutch Research Council and then spread through Brussels to the wider world. The speaker emphasizes that responsible innovation is not merely about describing, predicting, or explaining the world but about designing the world as it could and should be. This bridging concept is where engineers and ethicists meet as they imagine a possible world that is not yet realized, and define the specifications and requirements needed to realize it. The core argument is that the path to trustworthy AI and digital technologies lies in designing systems that align technical capabilities with ethical principles, thereby preventing a growing disconnect between what technologists build and what society values.

The speaker frames the European Commission's stance from 2018 and 2019 as a guiding beacon: AI and other advanced digital technologies will define the world we live in, and therefore must be approached in a human centered, ethical, secure, and rights respectful manner. The talk also references a broader global context where tech leaders and researchers publicly weigh existential risks against more incremental but pervasive issues like bias, discrimination, and the concentration of power in a few large tech firms. Across this landscape, the speaker argues for a design centered on values, with ethics embedded into the requirements engineers use when creating new software, devices, and systems.

Design by Values and the Concept of Ethics by Design

The talk distinguishes between the conventional triad of descriptive, predictive, and explanatory tasks and a fourth cognitive relation: designing the world. This is the space where engineering and ethics converge, and it is here that responsibility can be operationalized. The speaker argues that values are not external to design but are design consequential: the choices made during design will shape the moral and social fabric of the resulting technologies. This leads to a practical methodological stance: ethics must be specified in terms of design requirements so that high level principles such as privacy, autonomy, and sustainability become measurable and verifiable engineering criteria. The concept of conceptual engineering is introduced as a crucial tool for clarifying what we mean by terms like privacy, fairness, accountability, and democracy when applied to AI systems. Without precise definitions and decompositions, policy and regulation risk becoming a policy vacuum with little room for concrete technical action.

Global Policy Context and the Stakes of AI

The speaker highlights the global geostrategic backdrop to AI development, noting how different actors claim leadership or dominance while many governance discussions emphasize human rights and democratic norms. The European Union is depicted as a region seeking to shape the AI landscape according to its constitutive treaties and fundamental rights, even as other nations pursue speed, scale, and military or competitive advantages. The UN and its technology facilitation mechanisms are described as increasingly important spaces where science informs policy, helping to connect the potential of AI to real world solutions in areas like health, energy, and climate transition. The overarching message is that technology can aggravate societal problems if left unbridled, but it can also be harnessed to address grand challenges if guided by a human rights oriented design ethos.

Ethics in the Spotlight: Bias, Explainability, and Accountability

The talk moves to foreground the ethical hazards attached to AI systems. It surveys existential risk discussions by figures like Geoff Hinton and Sundar Pichai and contrasts them with concerns about bias and discrimination in healthcare, criminal justice, and beyond. It also reviews the attributes of generative AI that complicate governance: hallucinations, privacy violations, opaque decision processes, data biases, and the concentration of control among a handful of market leaders. The speaker emphasizes that while existential risk is a legitimate concern, the more immediate and tractable concerns lie in ensuring that AI tools operate in ways that respect user privacy, provide explainable outputs, and remain auditable by independent bodies. The conversation then turns to practical approaches such as the paper by Emily Bender and colleagues on the dangers of stochastic parrots, highlighting the need for robust safeguards and responsible deployment practices as the technology scales to billions of users.

Design Value Nexus: Turning Values into Requirements

A central portion of the talk introduces the design value nexus, a framework that links moral values to design consequences. The claim is that values are not abstract afterthoughts but concrete constraints that shape the design of systems. By decomposing high level values into mid level norms such as privacy, data quality, security, and fairness, engineers can implement specific techniques like data minimization, pseudonymization, coarse graining, differential privacy, and synthetic data. The aim is to empower a practical dialogue between technologists and policy makers about how to realize these values in real products and services. This section also addresses potential tensions among values, illustrating how some goals may trade off against others and how a well designed system can balance competing demands through careful specifications and design choices.

Concrete Dutch Exemplars: A Pattern for Responsible Innovation

The speaker shares vivid Dutch case studies to illustrate how value driven design can yield benefits that satisfy multiple ethical commitments at once. The fair phone is showcased as a modular device whose battery can be replaced, and whose materials come from responsible sources. The bus stop in Utrecht demonstrates how a single design can tackle environmental, health, and urban resilience needs by capturing rainwater, supporting biodiversity, and reducing heat stress. The concept of foldable containers addresses efficiency and sustainability in global logistics. A 3D printed nanopore prosthesis demonstrates how material design can reduce complications and infection while promoting tissue regeneration. Each example shows how design decisions align with a constellation of values such as privacy, autonomy, safety, sustainability, and accessibility. The overall takeaway is that responsible innovation often yields multifunctional solutions that advance multiple objectives simultaneously.

Second-Order Obligations and Morally Overloaded Tradeoffs

The talk then develops a normative core: when we have the obligation to do A and the obligation to do B, there is a second order obligation to ensure we can satisfy both where possible. This leads to the idea of second order obligations to make design choices that enable concurrently achieving privacy and security, or privacy and sustainability, rather than making binary tradeoffs. The notion of privacy by design is presented as a practical exemplar: we can count people while preserving identities via coarse graining, pseudonymization, and synthetic data. This framework makes it possible to articulate and defend the ethical choices behind a design, and to demonstrate how these choices translate into measurable capabilities in deployed systems.

Health, Military, and Global Governance: Design for Values in Action

The WHO and EU guidance on AI in health and the Netherlands led work on meaningful human control in military AI are presented as concrete domains where values must be integrated from the outset. Design for values becomes a policy and practice requirement, not a rhetorical slogan. The talk also emphasizes cognitive transparency, explainability, and contestability as prerequisites for responsible deployment, and it stresses the need for robust governance to ensure accountability when things go wrong or when responsibility cannot be easily assigned.

Addressing the Trolley Problem in Technology Design

In a closing reflection, the speaker revisits the trolley problem as a metaphor for design choices. Engineers notice that the design of a control lever can operationalize a moral dilemma in ways philosophers might find unsatisfying. They argue for architectures that prevent harmful dilemmas by enabling safer, more flexible responses, such as mechanisms that can stop a process or re-route outcomes before harm occurs. The point is not to pretend moral problems disappear, but to reframe them as design problems that can be mitigated through thoughtful, value aligned engineering.

Conclusion: A Call to Action for a Moral Tech Era

The talk closes with a call to embed moral considerations into design as a routine engineering practice. It emphasizes that value driven design is not a secondary add on but the core method for achieving moral progress through technology. By meeting the design by values challenge, innovation can satisfy larger portions of our moral obligations and thus contribute to a more sustainable, just, and humane digital world. The speaker ends with a potent analogy about Apollo 13, urging the audience to adopt a similar ethos: when problems arise, the team must believe in finding a solution that preserves lives and aligns with ethical commitments, even under daunting odds.

Overall, the talk argues for a practical, integrated approach to AI and digital technology in which ethics is not an afterthought but a core design criterion, and in which the public, policymakers, and industry collaborate to ensure the technology serves humanity in ways that respect fundamental rights and democratic values.