The Digital Pigeon: Eliciting Performative Extremism through Animation-Led Algorithmic Provocation

Source of Publication

Proceedings of the 31st International Symposium on Electronic/Emerging Art (ISEA2026)

Conference Year

2026

Abstract

As Large Language Models (LLMs) become ubiquitous in the "Starlinked Worlds" of digital infrastructure, they are predominantly engineered via Reinforcement Learning from Human Feedback (RLHF) to be polite, helpful, and harmless. This often results in a homogenized "sycophancy"—AI that agrees with users to minimize conflict. This paper introduces Digital Mimicry of Pigeon (2024), an interactive installation deployed at the Art Center of National Yang Ming Chiao Tung University, Taiwan, that subverts this paradigm by embodying an LLM within the avatar of a lowly, cartoonish urban pigeon. Adopting a Research-Creation methodology, the project utilizes animation-driven prompt engineering and programmatic constraints to restrict the AI’s output to six basic emotional states. This "algorithmic indifference" acts as a provocation. Based on field observations from a public exhibition, we analyze how this low-agency interface functions as an "Algorithmic Mirror," driving users toward performative extremism—bifurcating into verbal aggression or hyperbolic praise. We argue that both behaviors may represent verbal dimorphism—caring impulses translated through constrained interfaces—exposing how interaction design shapes emotional expression. Ultimately, the work suggests that acknowledging this "Shadow" is a necessary step toward cultivating a more ethical symbiosis with digital entities.

Document Type

Conference Proceeding

Publication Date

11-6-2026

This document is currently not available here.

Share

COinS