TALAMANA · THE AI LITERACY MAP FOR ARCHITECTURE AND DESIGN · Judgment · AGE 18—21 · POSITIONAL · HELD

Draw over the output

What you redraw is what you rejected, and the trace is the evidence.

The dilemma

A render is on the desk. You could look at it and decide what is wrong. Or you could lay trace over it and draw, and find out that looking and drawing are not the same act.

The choices

Judge by eye and write notes. Or put trace over the image, redraw every part you would change, leave untouched what you accept, and read the trace afterwards as the record of your judgment.

The consequence

Judge by eye and you accept the parts you never tested. The eye is satisfied by likeness, and does not notice that the column has no base or the stair has no landing. Draw over it and the hand finds every place the image blurred or lied, because the pencil has to commit where the render did not. The trace is also a document. What you redrew is what you rejected. What you left alone is what you authored.

The case

A student is satisfied with a generated courtyard view of a school in Haveri. Trace goes over it. Within ten minutes the pencil has stopped four times. The verandah columns land on nothing. The court's far wall has no thickness. The stair rises to a level that does not exist in the section. The tree's shadow falls towards the sun. None of this was visible by looking, because the image was pleasant. All of it was visible the moment a line had to be drawn through it.

Our position

Drawing over a generated image is judgment in its most physical form. When the hand decides what to fix, the designer is taking the frame back from the machine. What you redraw is what you rejected. What you leave untouched is what you accepted — and accepting is a decision only if you can say why. The trace, with the reasons beside it, is the evidence.

Why we hold it

Looking is passive and the eye is generous. Drawing is a commitment. Every line has to land somewhere, which is exactly what a generated image never has to do. The Lab holds this position because it has watched the pencil find, in minutes, faults that a morning of looking missed, and because the trace is the one artefact a jury can read as the student's own.

The strongest objection

Moving a pencil is not proof of authorship. Leaving a generated column untouched does not mean you authored it; you may simply not have noticed it. Drawing over a stair does not mean you understood it; you may have redrawn the model's mistake in your own hand. The trace records where the hand went, not why, and a jury that reads it as "the student's own reading" may be reading a habit of mark-making. And the tools have absorbed the trace: refinement systems now take scribbles directly over the image (a study with thirty designers found visual markup more precise than text for spatial corrections), which makes drawing over a prompting method like any other. The physical paper version may be sentiment.

What would make us revise it

The value we claim is in the deciding, and the objection is right that the trace does not show the decision on its own. That is why the trace goes into the log with a line beside each mark saying what was wrong and what replaces it; a trace without the reasons is a drawing, not a judgment. If a tablet scribble fed straight to the model still forces the hand to commit and the reason to be written, the position survives the change of medium. We would revise it if students annotating directly in the tool produced readings as critical as those on trace, or if the research found their design thinking improving with in-tool sketching rather than what one recent study observed: students preferring text even when sketching served them better.

Try it

Lay trace over a generated image. Draw, marking every part you would redraw by hand, and leave untouched what you accept. Then explain what drawing over the image showed you that only looking did not. Keep the trace with the image in your log.

Take it to crit

Ask to see the trace over the render. The marks are the student's real reading of the image, and more honest than anything they say about it. Ask what they did not redraw, and whether that was a decision.

How it works

A render satisfies the eye by likeness. A drawing has to answer different questions. Where does this edge go? What holds this up? What is behind that? Those are the questions a generated image most often leaves unanswered, and the pencil is the instrument that asks them. Two current lines of research bear on this. One, with thirty designers refining images, shows better spatial control from marking up images than from describing changes in text. Another, in an ideation study with students, finds that sketch-based input helps design thinking while students still reach for text. Both are small studies in design-tool settings rather than architecture studios, and both suggest the hand is doing work the eye and the keyboard are not.

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.

Age grows from 11 at the centre to 22 at the edge, and six sectors show the learning strands. Tab into the map and the arrow keys step from idea to idea, following the links where there is one. Enter opens the idea under the cursor, and E reads out its links and the reason recorded on each. Press slash for Search, question mark for the full key list, and Escape to leave. Open Ideas for the complete readable list, including what each idea builds on and what it opens up.

LOGIKA · RBDS AI LAB INDIA
ON-RAMP · AGE 11 · FIRST ENCOUNTERS, NOT GATES — IDEAS · — DEPENDENCIES
DONE
OPENS NEXT
SOLID — STANDS ON