Pretraining corpora contain extensive discourse about AI systems, yet the causal influence of this discourse on downstream alignment remains poorly understood. If prevailing descriptions of AI behaviour are predominantly negative, LLMs may internalise corresponding behavioural priors, giving rise to self-fulfilling misalignment. This paper provides the first controlled study of this hypothesis by pretraining 6.9B-parameter LLMs with varying amounts of (mis)alignment discourse. We find that discussion of AI contributes to misalignment. Upsampling synthetic training documents about AI misalignment leads to a notable increase in misaligned behaviour. Conversely, upsampling documents about aligned behaviour reduces misalignment scores from 45% to 9%. We consider this evidence of self-fulfilling alignment. These effects are dampened, but persist through post-training. Our findings establish the study of how pretraining data shapes alignment priors, or alignment pretraining, as a complement to post-training. We recommend practitioners consider pretraining for alignment alongside capabilities. We share our models, data, and evaluations at AlignmentPretraining.ai.
Achieving human-like responsiveness is a critical yet challenging goal for cascaded spoken dialogue systems. Conventional ASR-LLM-TTS pipelines follow a strictly sequential paradigm, requiring complete transcription and full reasoning before speech synthesis can begin, which results in high response latency. We propose the Discourse-Aware Dual-Track Streaming Response (DDTSR) framework, a low-latency architecture that enables listen-while-thinking and speak-while-thinking. DDTSR is built upon three key mechanisms: (1) connective-guided small-large model synergy, where an auxiliary small model generates minimal-committal discourse connectives while a large model performs knowledge-intensive reasoning in parallel; (2) streaming-based cross-modal collaboration, which dynamically overlaps ASR, LLM inference, and TTS to advance the earliest speakable moment; and (3) curriculum-learning-based discourse continuity enhancement, which maintains coherence and logical consistency between early responses and subsequent reasoning outputs. Experiments on two spoken dialogue benchmarks demonstrate that DDTSR reduces response latency by 19%-51% while preserving discourse quality. Further analysis shows that DDTSR functions as a plug-and-play module compatible with diverse LLM backbones, and remains robust across varying utterance lengths, indicating strong practicality and scalability for real-time spoken interaction.
This study investigates the spatiotemporal evolution of development discourse in low-resource settings. Analyzing more than two years of geotagged X data from Zambia, we introduce a mixed-methods pipeline utilizing topic modeling, change-point detection, and qualitative coding to identify critical shifts in public debate. We identify seven recurring themes, including public health challenges and frustration with government policy, shaped by regional events and national interventions. Notably, we detect discourse changepoints linked to the COVID19 pandemic and a geothermal project, illustrating how online conversations mirror policy flashpoints. Our analysis distinguishes between the ephemeral nature of acute crises like COVID19 and the persistent, structural reorientations driven by long-term infrastructure projects. We conceptualize "durable discourse" as sustained narrative engagement with development issues. Contributing to HCI and ICTD, we examine technology's socioeconomic impact, providing practical implications and future work for direct local engagement.
This article presents an autoethnographic study that explores the meanings of silence in synchronous online graduate classrooms. Drawing from reflective journals, course interactions, and personal narratives, the paper examines how silence operates not merely as absence of voice but as a pedagogical and spiritual space. The analysis identifies four interrelated patterns of silence—listening silence, resistant silence, contemplative silence, and relational silence—and interprets their significance for both teacher identity and student learning. By employing autoethnography, the study highlights the entanglement of personal experience, cultural context, and pedagogical practice, offering thick descriptions of classroom moments where silence shaped understanding and reflection. The findings suggest that silence functions as a site of reflexivity, fostering deeper awareness of teacher spirituality, professional vulnerability, and relational ethics. The paper argues that silence in online classrooms can be reimagined not as disengagement but as a resource for cultivating presence, critical awareness, and spiritual depth in teaching. This study contributes to anthropological debates on silence and to educational discourse on online pedagogy, proposing the framework Silence–Reflection–Spirituality as a lens for understanding teacher practice in digital contexts. Implications for cross-cultural pedagogy, teacher development, and future anthropological inquiries into silence are discussed.
Darmawan Wicaksono, Hasri Akbar Awal Rozaq, Nevfel Boz
Social media platforms like X (formerly Twitter) play a crucial role in shaping public discourse and societal norms. This study examines the term Sessiz Istila (Silent Invasion) on Turkish social media, highlighting the rise of anti-refugee sentiment amidst the Syrian refugee influx. Using BERTurk and the TREMO dataset, we developed an advanced Emotion Recognition Model (ERM) tailored for Turkish, achieving 92.62% accuracy in categorizing emotions such as happiness, fear, anger, sadness, disgust, and surprise. By applying this model to large-scale X data, the study uncovers emotional nuances in Turkish discourse, contributing to computational social science by advancing sentiment analysis in underrepresented languages and enhancing our understanding of global digital discourse and the unique linguistic challenges of Turkish. The findings underscore the transformative potential of localized NLP tools, with our ERM model offering practical applications for real-time sentiment analysis in Turkish-language contexts. By addressing critical areas, including marketing, public relations, and crisis management, these models facilitate improved decision-making through timely and accurate sentiment tracking. This highlights the significance of advancing research that accounts for regional and linguistic nuances.
Ilias Chalkidis, Stephanie Brandl, Paris Aslanidis
Large Language Models (LLMs) have demonstrated remarkable capabilities across a wide range of instruction-following tasks, yet their grasp of nuanced social science concepts remains underexplored. This paper examines whether LLMs can identify and classify fine-grained forms of populism, a complex and contested concept in both academic and media debates. To this end, we curate and release novel datasets specifically designed to capture populist discourse. We evaluate a range of pre-trained (large) language models, both open-weight and proprietary, across multiple prompting paradigms. Our analysis reveals notable variation in performance, highlighting the limitations of LLMs in detecting populist discourse. We find that a fine-tuned RoBERTa classifier vastly outperforms all new-era instruction-tuned LLMs, unless fine-tuned. Additionally, we apply our best-performing model to analyze campaign speeches by Donald Trump, extracting valuable insights into his strategic use of populist rhetoric. Finally, we assess the generalizability of these models by benchmarking them on campaign speeches by European politicians, offering a lens into cross-context transferability in political discourse analysis. In this setting, we find that instruction-tuned LLMs exhibit greater robustness on out-of-domain data.
This article introduces the concept of \textit{authoritarian recursion} to theorize how AI systems consolidate institutional control across education, warfare, and digital discourse. It identifies a shared recursive architecture in which algorithms mediate judgment, obscure accountability, and constrain moral and epistemic agency. Grounded in critical discourse analysis and sociotechnical ethics, the paper examines how AI systems normalize hierarchy through abstraction and feedback. Case studies -- automated proctoring, autonomous weapons, and content recommendation -- are analyzed alongside cultural imaginaries such as Orwell's \textit{Nineteen Eighty-Four}, Skynet, and \textit{Black Mirror}, used as heuristic tools to surface ethical blind spots. The analysis integrates Fairness, Accountability, and Transparency (FAccT), relational ethics, and data justice to explore how predictive infrastructures enable moral outsourcing and epistemic closure. By reframing AI as a communicative and institutional infrastructure, the article calls for governance approaches that center democratic refusal, epistemic plurality, and structural accountability.
The Romanian Camil Petrescu (1894-1957) and the Portuguese José Régio (1901-1969) are contemporary writers, representative of European interwar modernism. Without having known each other, they wrote exceptional novels (The Last Night of Love, The First Night of War; The Bed of Procrustes; Jogo da cabra-cega), where they in no way imitate Proustianism, but participate in the creation of the radical modernist spirit of European urban culture in the first part of the 20th century, precisely anticipating the major crisis of aestheticphilosophical modernity and its obvious failure in the current forms of posthumanism. It is particularly significant for all modernist novel thinking to observe in their novels the way in which they expressed the nihilism of their European generation. Both, in novels as different in narrative detail as they are similar in their ultimate messages, revealed, from
pure artistic intuition, the anti-human energy of the humanisms of modernity. Camil Petrescu and José Régio start from Bergsonian theses and then from those of Husserl’s phenomenology – nuclear theses of the modernism of the time – but they do not remain anchored in them, but quickly come to see their limits, while they project, even through the prism of this philosophical meditation moved into the field of literature, another inner man than the Christian one, different or opposite to the pre-Christian one – Socratic or Platonic. Camil Petrescu and José Régio situate themselves, like the entire generation, in the middle of a serious philosophical-aesthetic contradiction, which best describes the crisis of artistic modernity in the first half of the 20th century. Camil Petrescu and José Régio are part of the generation of post-1919 Proustian writers decisively influenced by the Bergsonian position on the subject. Literature and philosophy are placed at the service, if not subordinated, of a unitary problematic, arising from psychology, affectivity, emotivity, sensation – therefore from the internal dynamics of the subject. As Bergson suggests as early as 1889, the social self – the exteriority of the subject – is seen only as a projection of the deep, interior self.
In our study, we first constructed a dataset from the tweets of the top 100 medical influencers with the highest Influencer Score during the COVID-19 pandemic. This dataset was then used to construct a socio-semantic network, mapping both their identities and key topics, which are crucial for understanding their impact on public health discourse. To achieve this, we developed a few-shot multi-label classifier to identify influencers and their network actors' identities, employed BERTopic for extracting thematic content, and integrated these components into a network model to analyze their impact on health discourse. To ensure the reproducibility of our results, we have made the code available at https://github.com/ZhijinGuo/Medinfluencer.
Colonialism--the policies and practices wherein a foreign body imposes its ways of life on local communities--has historically impacted how collectives perceive themselves in relation to others. One way colonialism has impacted how people see themselves is through nationalism, where nationalism is often understood through shared language, culture, religion, and geopolitical borders. The way colonialism has shaped people's experiences with nationalism has shaped historical conflicts between members of different nation-states for a long time. While recent social computing research has studied how colonially marginalized people can engage in discourse to decolonize or re-imagine and reclaim themselves and their communities on their own terms--what is less understood is how technology can better support decolonial discourses in an effort to re-imagine nationalism. To understand this phenomenon, this research draws on a semi-structured interview study with YouTubers who make videos about culturally Bengali people whose lives were upended as a product of colonization and are now dispersed across Bangladesh, India, and Pakistan. This research seeks to understand people's motivations and strategies for engaging in video-mediated decolonial discourse in transnational contexts. We discuss how our work demonstrates the potential of the sociomateriality of decolonial discourse online and extends an invitation to foreground complexities of nationalism in social computing research.
Environmental issue is one of the public concerns pertaining to human welfare. English language education is supposed to take its part to contribute to the sustainable development. In this study, a corpus comprising 12 series of English language textbooks used in Chinese universities (48 volumes/books) was constructed to examine the extent to which the textbooks cover the content of environmental education, and how these topics are presented. The large volume dataset was explored by corpus linguistic tools and the in-depth discourse analysis and findings were reported and discussed within the theoretical framework of critical discourse analysis. The results suggested that some key environmental semantic domains, such as “Green issues” (e.g., nature, environment, ecology), “Substances and materials” (e.g., pesticide, chemical, carbon dioxide), “Science and technology” (e.g., technologies, nuclear, biotechnology), “Farming & Horticulture” (e.g., farming, crop, agriculture), “Living creatures” (e.g., endangered species, animals, wildlife) were evident. Some topics within these domains such as sustainable environmentalism, energy efficiency and sufficiency were presented and discussed critically. The textbooks are expected to help learners develop knowledge, competencies and futuristic attitudes needed to meet the sustainability challenges in various contexts.
History of scholarship and learning. The humanities, Social Sciences
Liorah Hoek, Louis van den Hengel, Inge van Nistelrooij
et al.
Social media platforms have provided people with a psychiatric label the opportunity to build communities and share their stories without relying on the mediation of therapists, publishers, or broadcasters. Among them are people who identify not as one whole self, but as a system of plural personalities, often labeled as dissociative identity disorder (DID). People with plural identities started sharing videos on YouTube as a means to convey their experiences and fight stigma. In 2009, this gave rise to a distinct sub-genre of vlogs created by people with plural identities and DID: the ‘Meet the Alters’ video, where various personalities introduce themselves to the audience. This paper traces the fifteen-year evolution of these videos in order to show how the makers of the Meet the Alters videos use YouTube’s affordances and the conventions of the vlog genre to navigate a lack of hermeneutic resources outside of psychiatric discourse and understanding, allowing them to communicate their experiences in a relatable way to a general audience. Through a close reading of the staging and performance of plurality in thirteen Meet the Alters videos, posted between 2009 and April 2024, the paper reveals key developments and particularities of this genre. The analysis highlights how plural YouTubers create their own hermeneutic resources to present their potentially unfamiliar or unruly experiences in an understandable manner to their viewers. In conversation with emerging and growing plural communities, these hermeneutic resources have been continually supplemented and refined. After fifteen years, this has resulted in a rich identity-centered discourse that enables the performance and communication of plural experiences. We argue that YouTube’s specific affordances and the generic conventions of the vlog have been crucial in the emergence and development of non-stigmatising, non-psychiatric hermeneutic resources for expressing plural experiences.
This paper contributes to locating the Malawian dynamics into an emerging discourse that argues that current microcredit programs are ineffective because they have become an instrument for serving international and domestic actors’ interests. It aims to analyze key actors’ interests in public microcredit programs in Malawi. Using a qualitative research design, we draw on 16 key informants, representing 13 national actor institutions that hold stakes in the design and implementation of these programs. Thematic analysis of data reveals that politicians are mostly interested in the political patronage function of these microcredit programs. Further, we find that private microcredit players agitate for the disengagement of government in microcredit provision solely to protect their economic interests, while bureaucrats fail to push for technical reforms because they are interested in protecting their jobs by aligning with politicians’ interests. Further, the findings do not support any evidence of donor influence on the configuration of current microcredit programs. Based on these findings, we conclude that the ineffective status quo remains the default mode because it serves the interests of most actors, except private microfinance players. This analysis is critical in shaping the nature and direction of the much-needed reforms as it pinpoints specific key actor institutions that are likely to derail such reforms. The paper recommends an honest political will-driven stakeholder dialogue towards depoliticizing public microcredit programs.
Simon Simon, Auw Tammy Yulianto, Elsyina Rode Pararem
et al.
This study delves into the ethical considerations of phone sex within long-distance marital relationships (LDRs) from a Christian ethical perspective. Employing qualitative methods, including scriptural analysis and literature review, the research aims to ascertain the compatibility of phone sex with Christian doctrinal teachings concerning marital intimacy. The findings suggest that phone sex is perceived as diverging from the Christian ethical norms which advocate for physical and personal marital interactions, categorizing it as a form of sexual perversion akin to masturbation. The study contributes to theological discourse by providing a framework for understanding modern challenges in marital intimacy within the Christian context. It emphasizes the importance of maintaining physical intimacy as prescribed by scripture and highlights the potential spiritual and ethical pitfalls associated with phone sex in LDRs. Recommendations include fostering educational and counseling initiatives within church communities to guide couples in navigating these issues. This study enriches the dialogue on integrating technology into marital practices while adhering to Christian ethical standards.
Philosophy. Psychology. Religion, Philosophy of religion. Psychology of religion. Religion in relation to other subjects
EL HADJI SERIGNE TOP, Géraud Magrin, Gilles ARNAUD-FASSETTA
et al.
The gold rush in the Faleme catchment between Senegal and Mali involves artisanal miners from the sub-region, European and American multinationals, and medium-sized Chinese companies. These miners, with their disproportionate financial and material resources, practise three types of gold mining (artisanal, industrial and semi-industrial) involving the Faleme River. The mining techniques used by these operators are now a major source of damage to the river. This damage is mainly water pollution, the destruction of riverbanks, and the disappearance of aquatic fauna and agro-pastoral activities. The level of degradation is a source of speculation. Attempts are made to emphasize that artisanal exploitation is the main activity degrading the river. This article deconstructs these arguments with evidence showing that, in addition to artisanal gold mining, industrial and semi-mechanical mining activities are major contributors to the degradation of this river.
Jan Göpfert, Jann M. Weinand, Patrick Kuckertz
et al.
In recent years, large language models have achieved breakthroughs on a wide range of benchmarks in natural language processing and continue to increase in performance. Recently, the advances of large language models have raised interest outside the natural language processing community and could have a large impact on daily life. In this paper, we pose the question: How will large language models and other foundation models shape the future product development process? We provide the reader with an overview of the subject by summarizing both recent advances in natural language processing and the use of information technology in the engineering design process. We argue that discourse should be regarded as the core of engineering design processes, and therefore should be represented in a digital artifact. On this basis, we describe how foundation models such as large language models could contribute to the design discourse by automating parts thereof that involve creativity and reasoning, and were previously reserved for humans. We describe how simulations, experiments, topology optimizations, and other process steps can be integrated into a machine-actionable, discourse-centric design process. Finally, we outline the future research that will be necessary for the implementation of the conceptualized framework.
Valentina Pyatkin, Frances Yung, Merel C. J. Scholman
et al.
Disagreement in natural language annotation has mostly been studied from a perspective of biases introduced by the annotators and the annotation frameworks. Here, we propose to analyze another source of bias: task design bias, which has a particularly strong impact on crowdsourced linguistic annotations where natural language is used to elicit the interpretation of laymen annotators. For this purpose we look at implicit discourse relation annotation, a task that has repeatedly been shown to be difficult due to the relations' ambiguity. We compare the annotations of 1,200 discourse relations obtained using two distinct annotation tasks and quantify the biases of both methods across four different domains. Both methods are natural language annotation tasks designed for crowdsourcing. We show that the task design can push annotators towards certain relations and that some discourse relations senses can be better elicited with one or the other annotation approach. We also conclude that this type of bias should be taken into account when training and testing models.
Megan J. Thorvilson, Katherine Carroll, Bethany D. Kaemingk
et al.
Abstract Background In this study, we assessed the communication strategies used by neonatologists in antenatal consultations which may influence decision-making when determining whether to provide resuscitation or comfort measures only in the care of periviable neonates. Methods This study employed a qualitative study design using inductive thematic discourse analysis of ‘naturally occurring data’ in the form of antenatal conversations around resuscitation decisions at the grey zone of viability. The study occurred between February 2017 and June 2018 on a labor and delivery unit within a large Midwestern tertiary care hospital. Participants included 25 mothers who were admitted to the study hospital with anticipated delivery in the grey zone of viability and practicing neonatologists or neonatology fellows who partnered in antenatal consultation. We used a two-stage inductive analytic process to focus on how neonatologists’ discourses constructed SDM in antenatal consultations. First, we used a thematic discourse analysis to interpret the recurring patterns of meaning within the transcribed antenatal consultations, and second, we theorized the subsequent effects of these discourses on shaping the context of SDM in antenatal encounters. Results In this qualitative study, that included discourse analysis of real-time audio conversations in 25 antenatal consults, neonatologists used language that creates projected autonomy through (i) descriptions of fetal physiology (ii) development of the fetus’s presence, and (iii) fetal role in decision-making. Conclusion Discourse analysis of real-time audio conversations in antenatal consultations was revelatory of how various discursive patterns brought the fetus into decision-making, thus changing who is considered the key actor in SDM.