Katherine Abramski

Giulio Rossetti, Massimo Stella, Rémy Cazabet et al.

In this paper we introduce Y, a new-generation digital twin designed to replicate an online social media platform. Digital twins are virtual replicas of physical systems that allow for advanced analyses and experimentation. In the case of social media, a digital twin such as Y provides a powerful tool for researchers to simulate and understand complex online interactions. {\tt Y} leverages state-of-the-art Large Language Models (LLMs) to replicate sophisticated agent behaviors, enabling accurate simulations of user interactions, content dissemination, and network dynamics. By integrating these aspects, Y offers valuable insights into user engagement, information spread, and the impact of platform policies. Moreover, the integration of LLMs allows Y to generate nuanced textual content and predict user responses, facilitating the study of emergent phenomena in online environments. To better characterize the proposed digital twin, in this paper we describe the rationale behind its implementation, provide examples of the analyses that can be performed on the data it enables to be generated, and discuss its relevance for multidisciplinary research.

Katherine Abramski, Giulio Rossetti, Massimo Stella

Implicit biases in both humans and large language models (LLMs) pose significant societal risks. Dual process theories propose that biases arise primarily from associative System 1 thinking, while deliberative System 2 thinking mitigates bias, but the cognitive mechanisms that give rise to this phenomenon remain poorly understood. To better understand what underlies this duality in humans, and possibly in LLMs, we model System 1 and System 2 thinking as semantic memory networks with distinct structures, built from comparable datasets generated by both humans and LLMs. We then investigate how these distinct semantic memory structures relate to implicit gender bias using network-based evaluation metrics. We find that semantic memory structures are irreducible only in humans, suggesting that LLMs lack certain types of human-like conceptual knowledge. Moreover, semantic memory structure relates consistently to implicit bias only in humans, with lower levels of bias in System~2 structures. These findings suggest that certain types of conceptual knowledge contribute to bias regulation in humans, but not in LLMs, highlighting fundamental differences between human and machine cognition.

Katherine Abramski, Riccardo Improta, Giulio Rossetti et al.

Free associations have been extensively used in cognitive psychology and linguistics for studying how conceptual knowledge is organized. Recently, the potential of applying a similar approach for investigating the knowledge encoded in LLMs has emerged, specifically as a method for investigating LLM biases. However, the absence of large-scale LLM-generated free association norms that are comparable with human-generated norms is an obstacle to this new research direction. To address this limitation, we create a new dataset of LLM-generated free association norms modeled after the "Small World of Words" (SWOW) human-generated norms consisting of approximately 12,000 cue words. We prompt three LLMs, namely Mistral, Llama3, and Haiku, with the same cues as those in the SWOW norms to generate three novel comparable datasets, the "LLM World of Words" (LWOW). Using both SWOW and LWOW norms, we construct cognitive network models of semantic memory that represent the conceptual knowledge possessed by humans and LLMs. We demonstrate how these datasets can be used for investigating implicit biases in humans and LLMs, such as the harmful gender stereotypes that are prevalent both in society and LLM outputs.

Virginia Morini, Valentina Pansanella, Katherine Abramski et al.

Social media platforms are online fora where users engage in discussions, share content, and build connections. This review explores the dynamics of social interactions, user-generated contents, and biases within the context of social media analysis (analyzing works that use the tools offered by complex network analysis and natural language processing) through the lens of three key points of view: online debates, online support, and human-AI interactions. On the one hand, we delineate the phenomenon of online debates, where polarization, misinformation, and echo chamber formation often proliferate, driven by algorithmic biases and extreme mechanisms of homophily. On the other hand, we explore the emergence of online support groups through users' self-disclosure and social support mechanisms. Online debates and support mechanisms present a duality of both perils and possibilities within social media; perils of segregated communities and polarized debates, and possibilities of empathy narratives and self-help groups. This dichotomy also extends to a third perspective: users' reliance on AI-generated content, such as the ones produced by Large Language Models, which can manifest both human biases hidden in training sets and non-human biases that emerge from their artificial neural architectures. Analyzing interdisciplinary approaches, we aim to deepen the understanding of the complex interplay between social interactions, user-generated content, and biases within the realm of social media ecosystems.

Katherine Abramski, Giulio Rossetti, Massimo Stella

As Large language models (LLMs) become increasingly integrated into our lives, their inherent social biases remain a pressing concern. Detecting and evaluating these biases can be challenging because they are often implicit rather than explicit in nature, so developing evaluation methods that assess the implicit knowledge representations of LLMs is essential. We present a novel word association network methodology for evaluating implicit biases in LLMs based on simulating semantic priming within LLM-generated word association networks. Our prompt-based approach taps into the implicit relational structures encoded in LLMs, providing both quantitative and qualitative assessments of bias. Unlike most prompt-based evaluation methods, our method enables direct comparisons between various LLMs and humans, providing a valuable point of reference and offering new insights into the alignment of LLMs with human cognition. To demonstrate the utility of our methodology, we apply it to both humans and several widely used LLMs to investigate social biases related to gender, religion, ethnicity, sexual orientation, and political party. Our results reveal both convergences and divergences between LLM and human biases, providing new perspectives on the potential risks of using LLMs. Our methodology contributes to a systematic, scalable, and generalizable framework for evaluating and comparing biases across multiple LLMs and humans, advancing the goal of transparent and socially responsible language technologies.

Katherine Abramski, Salvatore Citraro, Luigi Lombardi et al.

Large language models are becoming increasingly integrated into our lives. Hence, it is important to understand the biases present in their outputs in order to avoid perpetuating harmful stereotypes, which originate in our own flawed ways of thinking. This challenge requires developing new benchmarks and methods for quantifying affective and semantic bias, keeping in mind that LLMs act as psycho-social mirrors that reflect the views and tendencies that are prevalent in society. One such tendency that has harmful negative effects is the global phenomenon of anxiety toward math and STEM subjects. Here, we investigate perceptions of math and STEM fields provided by cutting-edge language models, namely GPT-3, Chat-GPT, and GPT-4, by applying an approach from network science and cognitive psychology. Specifically, we use behavioral forma mentis networks (BFMNs) to understand how these LLMs frame math and STEM disciplines in relation to other concepts. We use data obtained by probing the three LLMs in a language generation task that has previously been applied to humans. Our findings indicate that LLMs have an overall negative perception of math and STEM fields, with math being perceived most negatively. We observe significant differences across the three LLMs. We observe that newer versions (i.e. GPT-4) produce richer, more complex perceptions as well as less negative perceptions compared to older versions and N=159 high-school students. These findings suggest that advances in the architecture of LLMs may lead to increasingly less biased models that could even perhaps someday aid in reducing harmful stereotypes in society rather than perpetuating them.