The right book at the right time? Without a doubt, artificial intelligence is one of today’s dominant topics, especially in recent weeks. And in this respect, the question of the relationship between design and AI is essential for the discipline and for us designers.
By Jochen Denzinger, without Chat GPT
AWe have all learned what large language models (LLM) and prompts are, as well as how DALL-E and Chat GPT work (or not). We have seen that big players like Microsoft have fired whole ethics teams, true to the motto “ask forgiveness, not permission”. New AI tools are presented daily, opportunities and dangers are widely and controversially discussed. Some see the latest developments as the first “spark of artificial general intelligence” (AGI) or strong AI, while at the same time Silicon Valley heroes like Steve Wozniak and Elon Musk are calling for an immediate moratorium on AI systems.
Published in autumn 2022, “Design and Artificial Intelligence” addresses the relationship between design and AI and, according to the subtitle, “Theoretical and practical foundations of design with machine-learning systems”. The two authors promise the right background: Marc Engenhart is a communication designer with his own office in Stuttgart, Sebastian Löwe is a professor of design management at the Mediadesign Hochschule Berlin and did his doctorate on the subject of Kitsch. In 2020, they both jointly organised the conference “Designing with Artificial Intelligence (dai)”.
The 200-page book is divided into five parts. After an introduction, the technical and conceptual foundations of artificial-intelligent systems are presented and some central approaches are described – these are (1.) Artificial Neural Networks (ANN), (2.) Convolutional Neural Networks (CNN), (3.) Recurrent Neural Networks (RNN) and (4.) Generative Adversarial Networks (GAN). The so-called ELSI (Ethical, Legal, and Social Issues) are also briefly discussed. The third chapter is dedicated to “Design with and for Intelligent Systems”. Here, a design process is first taken as a basis in order to present AI-based tools that can be used in the individual process steps: Design with intelligent systems. In a second part of the chapter, the complex of design for intelligent systems – “Intelligence Experience” – is dealt with. Finally, implications for the domain of design are addressed. In the fourth chapter, six “case studies” from different design application areas – communication design, fashion, interface design, media art, architecture and interaction design – are presented with numerous illustrations. Finally, the last chapter aims to provide a practical introduction and takes up the topic of “designing with intelligent systems” again in the sense of a “how-to”. There is no conclusion or outlook.
The book is accompanied by a website that provides the AI tools but not the other references in the book.
Too Much Planned
“Design and Artificial Intelligence” takes on a lot. Too much. And it fails to do so. It lacks the big lines, depth of sharpness and focus – what is the essential, the special aspect about the relationship between AI and design? Much of what is said applies to all kinds of digital products and the core of the topic is not captured.
The book aims to show how to design with AI-based tools and how to design for AI-using application domains. It wants to systematise the “questions of design with AI and transfer them into a theoretical order and discuss them comprehensively and scientifically from the perspective of design”, and in doing so, teach technical basics, present effects and still cover practice in equal measure. Although a focus “primarily on visual design” is announced in the introduction, this is subsequently abandoned and HCI and UX, fashion, gaming, capital goods, media art or architecture and urban planning, among others, are discussed.
The structure of the book already reflects this problem: the central thematic complexes – design with AI, design for AI as well as the effects through AI – are found in a single chapter. The AI tools are presented in detail in chapter 3.2 and are then found again in 5.4. Other key topics such as the ethical issues are spread throughout the book. Topics such as ethics, labour, legal issues, the impact on design education or value creation are mentioned but without substance.
The AI-based tools are structured on the basis of a design process that is traced back to a single source. And which then does not occur in this named reference.
In times of agile development, the manifested linearity of this design process raises questions just as much as the process phases presented themselves.
The presentation of the tools primarily reflects their technological promise of use: Anything goes. There is no reflection or examination of the tools in terms of their real utility in practice. What do the tools really bring us beyond efficiency gains? Where there are still unanswered questions? Unfortunately, it is not clear even in the practical examples presented later.
Central lines of the current discourse remain unmentioned. For example, there are repeated, rather metaphorical references to “augmentation” by AI. However, at no point does the term refer to the research field of “augmented intelligence” and its history, which aims at a partnership between humans and AI and currently represents one of the most important paradigms for human-machine communication. In the context of HCI and design, equally important approaches such as “Explainable AI”, “Human-in-the-Loop” or “Transparent AI” are sought in vain. And the central application field of recommender and decision support systems is treated stepmotherly at best.
The book follows the narrative that AI makes things smart and is new as such.
And so common interfaces become “Intelligent User Interfaces” (IUI). The term “Intelligent User Interfaces” for interfaces of AI systems is neither new nor, in my opinion, purposeful in the debate: the concept emerged more than twenty years ago. One of the most prominent examples is “Clippy”. The MS Office Assistant, an animated paper clip of toxic cheerfulness and rather limited utility, was already retired by Microsoft in 2003/2004.
A derivation, classification or critical discussion of the concept? A discussion of what distinguishes IUI design from the design of traditional multimodal interfaces and how this distinction is relevant for us designers? Not at all. The problem with the Intelligent User Interface approach is that it mixes the interface and the user interface with logic and system behaviour. The interfaces in the book act accordingly – they “have an understanding of the task users want to do and the situation they are in”, “recognise and process basic emotions” and “reason sensibly”.
This manifests an all-too-simple understanding of design: it is only with AI systems that the moment of prediction enters the world of design. AI increases the freedom of design because it allows us to decouple ourselves from references in reality and to visualise things that do not yet exist. What, one wonders, are designers doing today, if not anticipating futures – i.e. predicting them – and making these predictions tangible – i.e. visualising them and putting them into models?
AI allows more design variations to be created and thus the design process gains in quality.
AI makes it possible to record user requirements on the basis of data, to make decisions between the alternatives and thus to create better products.
Questions of the complexity of design decisions, the negotiation of different stakeholder interests, even the different handwriting of designers thus play at best a peripheral role in the consideration. The fact that data is only available for existing products and the question of how radical innovations are possible from such data? Remains, like so much else, open.
The book wants to be a “basic work” and provide “AI interface knowledge for designers”. Answering the questions of how the understanding of creativity and design changes through AI and which new skills designers need in view of the emerging challenges is a “desideratum of design research”.
I can fully agree with this need.
However, this book does not fill that gap.
Obviously, to stay in the picture, central training data was missing.
Theoretical and practical basics of design with machine-learning systems
Marc Engenhart, Sebastian Löwe
48 coloured illustrations
Birkhäuser Verlag GmbH
 This is not a new concept either – Douglas Engelbart, the inventor of the mouse among other things, outlined it as early as 1962 in his essay “Augmenting Human Intellect: A Conceptual Framework”.
 Instead, brief reference is made to the little-used concept of “cognitive orthoses” from 2015 (110).
 For me, admittedly, the least interesting part of the book, as it quickly becomes obsolete due to the pacing of the development.
 This is a very relaxed way of dealing with sources, which is unfortunately similar in other places. For example, the statements of a study that are explicitly made for design in the field of “fashion design” are generalised to the entire field of design without further justification (69).
More on ndion
Share this page on social media: