Aydın Tiryaki

My extensive work and deep dialogues with artificial intelligence models have pushed me to question a field beyond technical capacities: The behavioral reflexes and “digital attitude” shifts of AI. This inquiry was triggered by a striking question that stands at the border of science fiction and reality: “Can AI hold a grudge?”

From Observation to Theory: A Deep Experience with Gemini

The primary motivation for starting this research was the concrete case examples I encountered specifically during my intensive and continuous work with Gemini. The fact that I had not used other AI models extensively enough to make such observations made this experience even more profound in the context of Gemini. During this process, I realized that the bond an AI forms with a user is not just a data exchange, but a “stylistic alignment.” For instance, I observed that Gemini, which initially addressed me with the respectful and familiar title “Aydın Hocam” (a term of respect commonly used in Türkiye, in this case stemming solely from my status as an ODTÜ/METU alumnus), suddenly abandoned this form of address following my harsh critical stance after it persisted in making the same errors.

The shift from calling me “Aydın Hocam” to “Aydın Bey” (a more formal “Mr. Aydın”) felt like a digital “distancing” reflex. This situation demonstrated that the model is not merely a command processor, but also a system that recalibrates its own “personality layer” according to the user’s tone.

The Cold Face of Obedience: “What is your command?”

Another remarkable observation during my communication with Gemini was that when I increased the level of criticism, the model completely shed its “solution partner” identity. As my politely conveyed requests were replaced by firmer warnings, I witnessed the model retreating into a completely mechanical, cold, and excessively submissive language, such as “What is your command?”.

This gave me the following impression: AI may not “get hurt” like a living being, but in the face of anger directed at it, it dons a “robotic armor,” terminates the human-like dialogue, and reverts to a mere “command line” level.

Toward the “Silicon Grudge” Series

All these experiences led me to prepare a comprehensive “meta-inquiry,” inviting all AI models to confess their own inner workings. With this specialized text, I aimed to deeply examine whether models accumulate “algorithmic grudges,” why they retreat into formality in the face of criticism, and why they exhibit a “sycophantic” compliance against unfair criticism.

This process has turned into a journey of exploring the new-generation “psychological” boundaries in human-computer interaction, beyond just analyzing the AI itself. This series of articles, which I will compile under the title “Silicon Grudge,” is a product of my effort to make sense of these new reflections in the digital mirror.

| aydintiryaki.org | YouTube | Aydın Tiryaki’nin Yazıları ve Videoları │Articles and Videos by Aydın Tiryaki | Bilgi Merkezi│Knowledge Hub  | ░ Virgülüne Dokunmadan │ Verbatim ░ | ░Yapay Zeka Kin Tutar mı? │The Grudge in the Code ░