arXiv subjects

Nico Curti

Publications and source records attributed to Nico Curti.

2 recordsLinked to original sources

arXiv2026

Tell me Mr. AI, what do you see in this image?

Background: The Rorschach inkblots are ambiguous stimuli developed to evoke subjective interpretations in humans, while modern artificial intelligence (AI) models are trained to recognize well established patterns and classes. The comparison of these two opposite systems arises a simple and provocative question: what happens when we ask an AI model to interpret an inkblot that "is not supposed to represent anything predefined"? Methods: We submitted the complete set of ten Rorschach inkblots to 61 AI models pretrained on the ImageNet dataset, spanning multiple architectural families. Model predictions were analyzed at the level of top-ranked classes and were quantified using a selected set of psycho-semantic variables inspired by the Rorschach tradition. Statistical analyses examined the effects of model family, computational complexity, and image conditions, comparing model-generated responses with human reference profiles. Findings: Across all architectures, model responses were highly non-random and showed systematic semantic convergence and inter-model agreement. However, quantitative analyses revealed a clear and robust separation between human responses and all AI model families. Human profiles exhibited substantially higher affective load, semantic richness, projected agency, and variability, whereas AI models converged toward frequent, formally coherent, and perceptually stable interpretations. Interpretation: Vision models consistently project the semantic organization learned, favoring consensus and formal coherence over affective or symbolic elaboration. Applying the Rorschach test to AI systems does not assess human-like cognition but provides a principled framework for exposing perceptual and semantic biases embedded in contemporary computer vision models.

physics.soc-ph↗

arXiv2026

On the Concept of Violence: A Comparative Study of Human and AI Judgments

Background: What counts as violence is neither self-evident nor universally agreed upon. While physical aggression is prototypical, contemporary societies increasingly debate whether exclusion, humiliation, online harassment or symbolic acts should be classified within the same moral category. At the same time, Large Language Models (LLMs) are being consulted in everyday contexts to interpret and label complex social behaviors. Whether these systems reproduce, reshape or simplify human conceptions of violence remains an open question. Methods: Here we present a systematic comparison between human judgements and LLM classifications across 22 scenarios carefully designed to be morally dividing, spanning from physical and verbally aggressive behavior, relational dynamics, marginalization, symbolic actions and verbal expressions. Human responses were compared with outputs from multiple instruction-tuned models of varying sizes and architectures. We conducted global, sentence-level and thematic-domain analyses, and examined variability across models to assess patterns of convergence and divergence. Findings: This study treats violence as a strategically chosen proxy through which broader belief formation dynamics can be observed. Violence is not the focus of the study, but it serves as a tool to investigate broader analysis. It enables a structured investigation of how LLMs operationalize ambiguous moral constructs, negotiate conceptual boundaries, and transform plural human interpretations into singular outputs. More broadly, the findings contribute to ongoing debates about the epistemic role of conversational AI in shaping everyday interpretations of harm, responsibility and social norms, highlighting the importance of transparency and critical engagement as these systems increasingly mediate public reasoning.

physics.soc-ph↗