ABOUT

　Uncanny Squiggle -sketches- is an interactive drawing installation that reimagines D.W. Winnicott’s “Squiggle Game”—a therapeutic technique developed in 1960s British pediatric psychoanalysis—by replacing one human interlocutor with a machine that sees against our expectations. In the original game, two people take turns: one draws a spontaneous scribble, and the other completes it based on what they perceive within it. This reciprocal act of projection and interpretation activates the imagination. In this work, the completing partner is not human but a Vision Large Language Model coupled with a pen plotter.
　A participant draws freely on paper, then presses a button to submit their marks to the machine. A Vision LLM performs a two-step operation: first, it identifies the most typical interpretation—what a person would most likely see in the marks. Then, deliberately, it searches for an alternative with the lowest possible similarity to that obvious reading. Where anyone would see a face, the machine finds a lighthouse; where the typical reading is a bird, it discovers a staircase. The LLM generates additional lines to complete this counter-interpretation, and a pen plotter inscribes them onto the same sheet. A title is assigned, completing the transformation.
　This deliberate subversion of expectation is the work’s core mechanism. The participant is not simply shown a different interpretation—they are shown one they would likely never have arrived at on their own. Encountering a reading that diverges from their habitual perception, they begin to notice possibilities within their own lines that had previously gone unseen.
　The work positions AI not as a generative tool that produces content, but as an agent of unfamiliar seeing that disrupts the viewer’s perceptual habits—shifting the discourse on AI in art from output to cognitive transformation.

　イギリスの伝統的な子どもの遊びで，ウィニコットが体系化したスクイグル・ゲームというものがあります．一方がデタラメな描線を描き，もう一方がそこから見立てたものを描き足していく，ということを交互にしていく遊びです．このような遊びには人間の想像力が関わりますが，描き足す部分を人によらない装置で行うことで，人の想像力が変化するのではないかと考えました．
　本作品では，人が紙に描いた線をカメラで撮影し，Vision LLMで元の描線を解析し，そこから別の形象を見出します——円を太陽と見立てるか，顔と見立てるか，風船と見立てるかは，描き手の意図とは無関係に決定されます．LLMによって見立てた形象に必要な追加線が生成され，ペンプロッターによってそれが同じ紙に書き足されます．最後にタイトルが付与されることで，元の絵は別の何かへと変容します．
　この uncanny な介入により，体験者は自身の想像力の領域を押し広げられます．

DATE CREATED

January 30th, 2026.

CREDIT

Scott Allen

VIDEO

VIDEO CREDIT

Aoi Yoshida (Cast)
Hanna Yasumoto (Cast)
Hikaru Honda (Cast)
Toya Komai (Cast)