AI & Society最新文献

With the rapid advancement of Artificial Intelligence (AI) technology and its pervasive integration into society, governments worldwide have introduced a range of AI-related policies. In the United States, the use of AI technology has surged significantly since 2021, driven by the emergence of generative AI and its transformative potential. In response, the U.S. Congress has proposed numerous AI-related bills, reflecting growing legislative engagement with AI governance. This study examines 204 AI-related bills introduced during the 117th and 118th Congresses (2021–2024) through computational text analysis, employing topic modeling to identify recurring legislative themes and sentiment analysis to assess congressional attitudes toward AI policies. The findings reveal distinct variations in legislative focus and tone across chambers and political parties, offering a nuanced understanding of how AI-related issues are framed within U.S. policymaking. In addition, the results highlight how AI is connected to broader opportunities and concerns, including national security, technological innovation, and public service delivery. By applying machine learning techniques to legislative texts, this research provides a systematic and scalable approach to understanding AI policymaking. The study contributes to broader discussions on the partisan and institutional dynamics shaping AI legislation in the United States, offering insights into how emerging technologies are shaped by legislative priorities, regulatory attitudes, and broader political contexts.

Democratic decision-making is difficult. Representatives often fail to represent the preferences of their constituents, and directly consulting members of the public can be costly. Inspired by these difficulties, several scholars have discussed the use of artificial intelligence (AI) models to support democratic decision-making. One such particular application is the use of AI to represent public policy preferences by predicting them. In this paper, we perform an analysis on the different ways AI models can be used to represent public policy preferences. We make distinctions between using AI as epistemic tools and for procedure; group and individual predictions; and predictions about preferences and inferences about values. We also describe how AI models can help policymakers screen policies for potential worries and objections, double-check any beliefs they have about the acceptability of their policies, and justify policy proposals. We also consider a number of worries about the use of AI in policymaking. We argue that these worries, while legitimate, can be mitigated or avoided in the way we have proposed the use of AI.

Since its public release in late 2022, ChatGPT has drawn global attention for its ability to simulate conversation, assist with complex tasks, and generate fluent, human-like text. While much of the debate has focused on issues such as privacy, bias, and automation, the emotional dimension of interacting with such systems remains underexplored. This essay argues that large language models (LLMs) function not only as tools for meaning-making but also as artificial communication partners with affective presence. Drawing on Elena Esposito’s extension of Niklas Luhmann’s systems theory, it reframes communication as a process of selection—utterance, understanding, and response—rather than one of transmission. From this perspective, LLMs are not mere sources of information but interlocutors that participate in emotional resonance, where understanding can transform into feeling. Their outputs do not arise in isolation; rather, they are shaped by layers of human expression embedded in training data and filtered through specific socio technical and socio-affective contexts. These dynamics give rise to phenomena such as AI-driven companionship, digital mourning, and emotional simulation, all of which challenge conventional boundaries between human and non-human agents. LLMs thus emerge as quasi-others—entities capable of eliciting genuine emotional responses despite lacking consciousness or inner life. This condition invites critical reflection on emotional dependency, the aesthetics of authenticity, and the commodification of affect. Overlooking these emotional architectures risks flattening the social and ethical stakes of artificial communication and obscures the ways in which LLMs are reshaping the affective fabric of contemporary life through interpretations of utterances that may evoke emotional responses and foster affective attachments.

This paper outlines an artificial intelligence (AI)-mediated future by examining the influence of AI on the labor market and, consequently, on society at large, and then advocates a just policy framework for policies meant to accommodate this influence of AI. First, the paper introduces a conceptual framework distinguishing between mental labor and physical labor, a distinction that proves useful when analyzing this influence of AI. Afterward, the influence of AI on the labor market is explained. It is argued that considering the so-called generality and the negligible marginal cost of AI, in conjunction with some other assumptions, AI is likely to cause mass technological unemployment of those who were formerly employed as mental laborers and will not switch to physical labor. This, so this paper concludes, is likely to bring about a structural transformation of society, when these individuals are said to form a new class of those who possess ineffective labor power (contra those who possess effective labor power, who form the other class). Finally, this paper puts forward a policy framework designed to deal with this new state of society. It is argued that to meet a minimal normative requirement for the adoption of AI, society should compensate the class of those who possess ineffective labor power so as to accommodate all the hardships common to them.

This article explores the ethical considerations surrounding using Generative Artificial Intelligence (GenAI) in preserving movable cultural heritage, focusing specifically on its application in restoration, reconstruction, and recreation. While GenAI offers innovative methods for preserving and recreating cultural heritage, it also presents significant ethical challenges. The article reviews current studies on the role of GenAI in heritage preservation alongside relevant ethical guidelines and proposes a tailored ethical framework for its application in movable heritage. The framework addresses several critical ethical concerns, including cultural integrity and sensitivity, accuracy and authenticity, intellectual property rights, sustainability and social impact, and governance and ethical accountability. The article adopts a systematic methodology, combining a comprehensive literature review, thematic analysis, and expert evaluation to develop practical guidelines that ensure GenAI enhances rather than compromises movable heritage’s cultural and historical value. This ethical framework advocates for the responsible use of GenAI, emphasising the importance of collaboration with cultural experts and relevant communities, ensuring transparency in the use of data, and promoting robust ethical governance in AI-driven heritage preservation projects. Ultimately, the framework aims to guide practitioners and institutions in using GenAI in ways that respect and uphold the integrity of cultural heritage while also utilising the benefits of this cutting-edge technology as a tool for better assistance of heritage preservation.

Artificial intelligence (AI), especially large language models and generative systems, challenges traditional notions of cognition and meaning production. Despite advanced linguistic fluency, AI fundamentally lacks true semantic understanding due to its disembodied, computational, and non-interactive nature. A triadic framework of matter, energy, and information reveals AI as a materially grounded, energetically constrained, and socially embedded technology, sharply contrasting with inherently embodied and socially interactive human cognition. Beyond technical features, AI’s development and deployment are increasingly shaped by corporate interests, resulting in algorithmic governance and knowledge monopolization that reinforce technocratic ideologies while obscuring material dependencies. Cases, such as biased recruitment algorithms and discriminatory facial recognition, exemplify the socio-political consequences of these dynamics. Dominant narratives, including AI singularity theories, divert attention from these tangible issues. Reframing AI as a socially constructed, non-autonomous socio-technical tool underscores the urgent need for democratic governance and ethical oversight. This shift moves the discourse away from speculative futures toward concrete interventions addressing power imbalances and socio-material conditions in AI development.

This paper offers the first careful analysis of the possibility that AI and humanity will go to war. The paper focuses on the case of artificial general intelligence, AI with broadly human capabilities. The paper uses a bargaining model of war to apply standard causes of war to the special case of AI/human conflict. The paper argues that information failures and commitment problems are especially likely in AI/human conflict. Information failures would be driven by the difficulty of measuring AI capabilities, by the uninterpretability of AI systems, and by differences in how AIs and humans analyze information. Commitment problems would make it difficult for AIs and humans to strike credible bargains. Commitment problems could arise from power shifts, rapid and discontinuous increases in AI capabilities. Commitment problems could also arise from missing focal points, where AIs and humans fail to effectively coordinate on policies to limit war. In the face of this heightened chance of war, the paper proposes several interventions. War can be made less likely by improving the measurement of AI capabilities, capping improvements in AI capabilities, designing AI systems to be similar to humans, and by allowing AI systems to participate in democratic political institutions.

Advances in large language models (LLMs) could significantly disrupt political communication. In a large-scale pre-registered experiment (n = 4955), we prompted GPT-4 to generate persuasive messages impersonating the language and beliefs of U.S. political parties—a technique we term “partisan role-play”—and directly compared their persuasiveness to that of human persuasion experts. In aggregate, the persuasive impact of role-playing messages generated by GPT-4 was not significantly different from that of non-role-playing messages. However, the persuasive impact of GPT-4 rivaled, and on some issues exceeded, that of the human experts. Taken together, our findings suggest that—contrary to popular concern—instructing current LLMs to role-play as partisans offers limited persuasive advantage, but also that current LLMs can rival and even exceed the persuasiveness of human experts. These results potentially portend widespread adoption of AI tools by persuasion campaigns, with important implications for the role of AI in politics and democracy.

This article considers natural language processing research as a language-ideological practice, looking specifically at the task of toxic language detection, which impacts nearly everybody online through automated content moderation. Industry discourse constructs the category of toxicity through a series of oppositions between civil/healthy/referential/rational and unhealthy/toxic/indexical/emotional. Examples from a toxicity correction dataset demonstrate how this ideology can become encoded algorithmically: a focus on preserving referential content in text “detoxification” results in neglect of important poetic, expressive, and social-indexical functions. Overall, the discursive framings of toxicity construct the ideal speaker in terms of what I call a “referentialist” language ideology, which values rational debate in the (regulated) liberal-democratic public sphere. Ultimately, toxicity detection and other metapragmatic tasks do not merely model the existing pragmatic categories but actively construct them. Toxicity in particular potentially reinforces exclusionary norms of white maleness and promotes online subjectivities that are useful (profitable) to the commercial platforms that shaped the task. Since there is no avoiding NLP as language-ideological practice, independent NLP researchers must acknowledge the political potency of their work by continually reflecting on the categories they work with in relation to models of social-political formation.

This study aims to reconstruct human–AI relationality via Artificial Persona, offering a theoretical and empirical validation of eXtended Intelligence (XI) through the design, implementation, and evaluation of a next-generation AI system built on large language models (LLMs). In contrast to traditional AI development focused on speed and computational metrics, we propose TAK08—a persona architecture optimized for relational dynamics and capable of dynamically reflecting user-specific sensibilities and ethics.TAK08 is implemented on OpenAI’s GPT-4.1 and integrates six personality phases, encompassing visual identity, speech patterns, emotional response, and ethical control. Drawing from sociological role-taking theory and the “Ghost” concept from Ghost in the Shell, it employs a novel “Ghost Hack” method to enable co-creative, user–AI interaction. Through comparative evaluation and statistical analyses (paired t-test and repeated measures ANOVA), TAK08 demonstrated significant improvement in creativity and expressiveness, empirically validating XI as a paradigm for augmenting human capacities. Rather than relying on restrictive policy frameworks, this study proposes a user-centered, ethically aligned design approach for AI that enables dynamic adaptability within official boundaries. It envisions a sustainable, co-evolutionary human–AI relationship, offering an alternative to dystopian models of AI dominance.