In the Trace of Water
TR

Istanbul Museum of Water Civilizations × Işık University · March – June 2026

Workshop and process

Six steps, four months. A process that began with analogue observation, continued with a model trained on the students' own vocabulary, and ended by comparing two design routes.

Where, and how, does water want to exist in this space?

Process 1 · AI as analyst

The machine reads, the student designs. A model trained on the students' photographs scores the route photo against the concepts, and the student uses those numbers as input.

Process 2 · AI as generator

The student writes, the machine draws. Three concepts become one design sentence, and a prompt generates an installation on top of a photograph of the real space.

The template given to students

“At the intersection of [Word 1] and [Word 2], [Word 3] becomes a spatial experience.”

Timeline

  1. 11 March 2026

    01 Brief

    The workshop guide was published and three phases were defined: observation and documentation, two design processes with AI, and comparison and discussion.

  2. 12–14 March 2026

    02 Observation

    Students searched the museum and its grounds for traces of water: architectural elements, surface materials, reflections of light, damp, sound, and images that represent water. Each concept needed at least 15 photographs, plus a 10-second video along a chosen route. This phase was fully analogue; the camera was the only tool.

  3. March 2026

    03 Concept set

    The students wrote the concept set themselves: thirty-two concepts, with no ready-made label list, all chosen by the group.

  4. 31 March – 2 April 2026

    04 Process 2 · AI as generator

    A bilingual prompt guide was handed out. Each student merged their three concepts into one design sentence, uploaded their route photo to Gemini and wrote a prompt asking it to place an installation in the real space.

  5. 12 April 2026

    05 Model

    Each concept became a class in Google Teachable Machine, and the students' own photographs became its training data. The result is a small image classifier that runs in the browser and scores any new photograph of the museum against the group's concepts.

  6. 19 May – 4 June 2026

    06 Process 1 · AI as analyst

    Students ran their route photos through the model. It scored every concept; the top three went to a language model, which wrote a one-sentence installation proposal for each, and a wireframe changed in proportion to the probabilities. Students used this reading as a measured input, not as an answer.

  7. June 2026

    07 Compare and submit

    Each student submitted their three concepts, their prompt, the AI-generated images and their own proposal for the same point on the route. The closing discussion asked whether water can exist in a space without being seen, and where AI belongs in the design process.

Thirty-two concepts

The students wrote the concept set themselves. No ready-made label list was used; the trained model is a record of how one group of students read one building.

Tools used

Closing questions

  1. Can water exist in a space even when it cannot be seen?
  2. At which step of the architectural design process should AI be used?
  3. How can human thinking and AI collaborate in producing design?

Team

The full team →

Project leads

  • Prof. Dr. Elif Süyük Makaklı
  • Dr. Lecturer Gülay Kepsutlu
  • Lecturer Dicle Zeycan

Curator-researchers

  • M.Arch. Sıla Yıldırım
  • M.Arch. Raşit Eren Cangür
  • M.Arch. Egemen Sezer
  • M.Arch. Eren Sezer