Reading Guide: Artificial Intelligence in Design '92

Long before generative design tools became a feature in everyday CAD software, researchers were asking how a computer could reason about design. Artificial Intelligence in Design '92, edited by John S. Gero with Fay Sudweeks, collects the papers from the Second International Conference on Artificial Intelligence in Design, held in Pittsburgh in June 1992. At more than 900 pages it is a large book, and this guide will help you find your way through it.

Get the book: View Artificial Intelligence in Design '92 on Amazon

What this book is (and who it's for)

This is a conference proceedings volume, first published by Kluwer Academic Publishers in 1992 and now part of Springer's catalogue (XI + 926 pages, 45 papers). The publisher presents it as a snapshot of the state of the art at the time, aimed at researchers, developers and users of computer systems for design, and as a book that shows both the breadth and the depth of AI in design.

The underlying argument is that design, in engineering and architecture alike, is a central driver of economic competitiveness and the step that comes before manufacturing, and that computational models built on the AI paradigm offer a new way to study and support it.

Who will get the most from it:

• Researchers in design computing tracing the roots of today's ideas in generative and knowledge-based design.
• Graduate students in engineering design, architecture or AI who need the historical literature for a thesis.
• Historians and teachers of AI interested in the knowledge-based systems era.

Some background in symbolic AI (search, rules, constraints, knowledge representation) and in design methods will help a lot. Readers trained only in modern deep learning should expect a different vocabulary.

Before you start

• Remember the context: this is pre-web, early-1990s research. Hardware was limited, and most systems were prototypes.
• Keep a short glossary as you read: terms like constraint propagation, feature-based modelling, shape grammars and design prototypes recur across papers.
• Have access to a library database so you can follow references to journals such as AI EDAM, which covered the same research community.

How to read it

Read it by theme rather than cover to cover. The publisher's contents list groups the papers into parts that include:

• Constraint Processes in Design — how design requirements can be represented as constraints and how systems satisfy or relax them. A good starting point for engineers, since the ideas map directly onto parametric CAD.
• Design Object Modeling — how to represent the things being designed: components, features and their relationships. For example, "Grammars of Features in Design" by K. N. Brown, J. H. Sims Williams and C. A. McMahon (pp. 287–306) looks at formal grammars for describing design features.
• Shape Design — computational approaches to form and geometry, of special interest to architects and product designers.

The book contains further themed sections beyond these. A suggested plan:

1. Skim the full table of contents and mark papers by your field: mechanical engineering, architecture, or AI methods.
2. Read one paper from each of the three parts above to get a feel for the range of approaches.
3. Go deep on your own theme, following references to later work by the same authors to see how their ideas developed.

On a first pass, skim implementation details of specific 1990s software systems; focus instead on the problem each paper sets out to solve and the representation it proposes.

Key ideas to take away

• Design as a reasoning process. The conference treated design not only as a product but as a process that can be modelled, studied and supported.
• Representation comes first. Much of the work is about how to describe design knowledge so a computer can use it.
• Constraints and grammars as generative tools. Both ideas live on in today's parametric and generative design software.
• Human-computer collaboration. The goal was usually to aid designers, not replace them, a framing that is newly relevant in the age of generative AI.

Practice and discussion questions

1. Choose one paper. Which of its ideas can you recognize in a modern CAD or BIM tool?
2. How did the researchers represent design knowledge, and what were the limits of that representation?
3. Compare a constraint-based approach with a grammar-based one. When would you choose each?
4. What problems did these systems struggle with that modern machine learning handles better, and the reverse?
5. How might a 1992 author react to today's text-to-3D or generative design tools?
6. Pick a design task from your own work. Sketch how you would formalize it as constraints or as a grammar.

Pair it with

• Parametric and generative design: the modern practice that grew partly out of this research.
• Knowledge representation and constraint satisfaction: the AI foundations that most papers rely on.
• Related books in our store: Artificial Intelligence: Its Scope and Limits offers a broader look at the AI ideas of that era, and An Artificial Intelligence Approach to Integrated Circuit Floorplanning shows AI applied to a specific engineering design problem. Browse them on the Books Plus Ai shelf.

Closing thoughts

Reading Artificial Intelligence in Design '92 is like visiting the workshop where many of today's design tools were first sketched out. Choose a theme, read a few papers carefully, and you will come away with a deeper understanding of why modern design software works the way it does. Get Artificial Intelligence in Design '92 on Amazon.

As an Amazon Associate, BooksPlusAi may earn from qualifying purchases.

Sources

• Springer: Artificial Intelligence in Design '92
• College of the Holy Cross Library catalogue record
• University College Cork: Grammars of Features in Design