Design Process of a Trusted AI

Design Process of a Trusted AI#

Designing a trusted AI system is a multidisciplinary challenge that requires integrating diverse tools, products, and services. Success depends on balancing performance, reliability, and aesthetics, while quickly translating scientific research into sustainable solutions. This process is fundamentally a design engineering problem.

Achieving an effective design process#

Products and services are the outputs of an organisation, and can be either tangible or intangible. Today, data-driven and non-deterministic models are increasingly embedded in many products. To design effectively, it is essential to understand the product’s nature and context, allowing for tailored approaches. Broadly, a “product” is any outcome of a process, and can be classified as:

  1. Hardware: Physical objects.

  2. Software: Intangible assets such as computer programs, laws, or policies.

  3. Services: Outcomes from interactions between suppliers and customers, including utilities, transportation, waste management, policing, retail, advertising, and information delivery.

  4. Processed materials: Bulk solids, liquids, or gases measured by volume, mass, or energy (e.g., plastic pellets, fuel, grease, coolant).

To achieve effective design, a designer should demonstrate:

  1. Practical competency: Applying experience, design principles, heuristics, and technical data.

  2. Subject competency: Understanding relevant fields, engineering sciences, and real-world examples.

  3. Method competency: Using design methods for analysis, experimentation, and creativity.

  4. Systems competency: Considering broader implications such as life cycle and economics.

  5. Social competency: Practicing teamwork, communication, leadership, and cultural awareness.

  6. Economic competency: Recognizing costs, returns, business, and political factors.

Design engineering in the context of AI spans both backend systems and frontend user experience and every tangible and intangible assets that affect these systems:

Design Aspect

Software & AI Engineering (Backend/Systems)

UX/UI & Product Design (Frontend/User-Facing)

Core Subject

Architecture, algorithms, infrastructure; focus on performance, scalability, reliability.

User interface and experience; focus on usability, accessibility, engagement.

Representation & Analysis

System diagrams, UML charts, database schemas; code as the ultimate representation.

Wireframes, mockups, prototypes; visualizing user journeys and product flow.

Design Process

Structured, iterative methods (Agile, Scrum, TDD); emphasis on clean, efficient code.

Empathetic, creative methods (Design Thinking, User-Centered Design); emphasis on user needs.

Resulting Properties

Internal properties: low latency, high throughput, data security.

External properties: intuitive interface, user retention, positive feedback.

Method of Study

Code reviews, performance profiling, algorithmic analysis, rigorous testing.

Usability testing, A/B testing, user interviews, design improvements.

Key Responsibilities

System stability, data integrity, security, compliance with standards.

User satisfaction, business goals, brand consistency.

Work Environment

IDEs, version control, collaboration platforms (GitHub, Jira).

Design software, workshops, brainstorming sessions.

Systematic design engineering uses theory strategically to guide the process, while methodical design engineering applies both established and innovative methods, blending theory, best practices, and intuition.

Information in design can relate to tangible items, processes, or cognitive concepts. Process objects are divided by purpose: (1) practical use—manufacturing, distribution, operation, or disposal; or (2) design—creating tangible or process objects [164].

Centrality of design engineering in AI development#

AI development sits at the intersection of multiple disciplines:

  • Socio-Cultural Axis: Human and societal context.

    • Humanities: Ethical and philosophical considerations—fairness, bias, accountability.

    • Social Sciences: Human-centric design and societal impact—user behavior, adoption, integration.

  • Techno-Scientific Axis: Technical and scientific foundations.

    • Engineering Sciences: Fundamental research—mathematics, statistics, computer science.

    • Technological Sphere: Applied technology—algorithms, frameworks, hardware.

  • Econo-Organizational Axis: Business and operational aspects.

    • Production/Consumption: Business case, monetization, ROI, market demand.

    • Manufacturing Practice: MLOps, system architecture, deployment, monitoring.

At the center is AI Development, integrating all axes:

  • Engineering sciences provide theoretical foundations.

  • Technology enables building and deployment.

  • Social sciences guide responsible, user-friendly design.

  • Humanities ensure ethical alignment.

  • Business defines goals and value.

  • Manufacturing practice turns models into production systems.

Foundational skills—language, graphics, mathematics, statistics, and computing—are essential throughout all stages, from data collection to deployment and monitoring. AI is not just technical; it requires a holistic understanding of technology, business, and society.

Evaluation and verification in design#

During design, evaluation can address:

  • How good is the solution? (overall or partial evaluation)

  • Does it meet the task definition?

  • Which solution is optimal?

  • What are the optimal property values?

The likelihood of errors during verification depends on task originality and complexity, the quality of the task definition, and the state of the designers.

Fundamental skills for good design engineering#

  1. Iterative thinking: Refining solutions through repeated cycles.

  2. Problem decomposition: Breaking complex problems into manageable parts.

  3. Abstraction: Focusing on essentials, deferring less critical details.

  4. Concretization: Turning abstract ideas into practical solutions.

  5. Analytical skills: Understanding causes, parameters, and system behaviors.

  6. Synthesis: Creatively combining elements for desired outcomes.

  7. Continuous improvement: Evaluating and enhancing designs.

  8. Strategic problem-solving: Addressing root causes, adapting approaches as needed.

Reading Materials#

This chapter draws extensively from the following reference:

  1. Eder, W. Ernst, and Hosnedl, Stanislav. Design Engineering: A Manual for Enhanced Creativity. United States, CRC Press, 2007. (https://www.google.co.uk/books/edition/Design_Engineering/RzIRTbZ-U6MC?hl=en&gbpv=1&pg=PP1&printsec=frontcover)

Contents