TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · Studio Practice · AGE 18—20 · FACTUAL · EVOLVING
AI and the material choice
The model narrows the options, and you decide with a sample in your hand.
When to use
When a material decision has more than two options and each comes with documents: a datasheet, a product declaration, a test certificate, a supplier's claim.
The method
One: gather the documents, not the adjectives — the EPD, the datasheet, the test report. Two: give the model the documents and ask for a table: embodied carbon with its boundary, the declared unit, strength or U-value, the distance from the named plant to your site, and the date of the declaration. Three: ask it to flag what each document does not say. Four: use the table to narrow to two or three options. Five: decide using what the table cannot hold — a sample in your hand, a visit to the kiln or the yard, the mason's word on what can be laid well here, and what the local supply will actually deliver in the monsoon. The model narrows. You decide.
Watch for this
A comparison across declarations with different boundaries or units. One EPD counts cradle-to-gate per tonne; another counts cradle-to-grave per square metre. A model will put them in one column and not notice. Make it state boundary and unit in every row.
Try it
Find two EPDs for a material you use — two bricks, two cements, two tiles — from the Indian or the international EPD library. Ask a model to build the comparison table with boundary, unit and date in every row. Then find the one line where it quietly compared unlike with unlike.
Prove it
Show one material decision with the model's comparison table beside it, the two lines that ruled options out, and the reason the final choice came from something the table could not hold.
How it works
An Environmental Product Declaration is an independently verified, registered document that reports a product's environmental performance from a life-cycle assessment, under rules written for its product category. EPD International runs the global programme, and the Indian EPD programme is aligned with it. These documents are long, structured and numerical — exactly what a language model reads well. What the model cannot help with is the part of a material decision that has no document: how a stone weathers on this coast, which kiln near Dharwad fires evenly, whether the mason has laid a cavity wall before. The method ends with those on purpose. Declarations are being issued faster than they are read, and Indian product coverage is growing from a small base. That is why this card is marked EVOLVING.
What this idea builds on
What this idea opens up
Sources
Open this idea on the map · The complete map · Logika · RBDS AI Lab, India · revised every edition.