<?xml version="1.0" encoding="utf-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
    <channel>
        <title>Discourses on Culture Feed</title>
        <link>https://reference-global.com/journal/DOC</link>
        <description>Reference Global RSS Feed for Discourses on Culture</description>
        <lastBuildDate>Thu, 22 Oct 2026 00:54:10 GMT</lastBuildDate>
        <docs>https://validator.w3.org/feed/docs/rss2.html</docs>
        <generator>https://github.com/jpmonette/feed</generator>
        <image>
            <title>Discourses on Culture Feed</title>
            <url>https://paradigm-files.s3.eu-central-1.amazonaws.com/DOC/cover-image.jpg</url>
            <link>https://reference-global.com/journal/DOC</link>
        </image>
        <copyright>All rights reserved 2026, SAN University</copyright>
        <atom:link href="https://reference-global.com/2/rss/journal/DOC" rel="self" type="application/rss+xml"/>
        <item>
            <title><![CDATA[Preface: Algorithm and Author: Retaining the Human Touch?]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2026-0001</link>
            <guid isPermaLink="false">/article/10.2478/doc-2026-0001</guid>
            <pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Visual Politics in the Age of Algorithms: A Four -Dimensional Model of Meaning, Visibility, Affect, and Legitimacy]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2026-0011</link>
            <guid isPermaLink="false">/article/10.2478/doc-2026-0011</guid>
            <pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[This article develops a four -dimensional conceptual framework for analysing visual politics under conditions of algorithmic mediation. It argues that contemporary political images become politically consequential through the interaction of four dimensions that exceed representational analysis alone: constructive meaning -making, regimes of visibility, affective mobilisation, and the visual performance of legitimacy. By synthesising insights from visual culture, political aesthetics, and international relations, the article reframes visual politics as a discursive and infrastructural process shaped by hybrid human -machine assemblages. It further argues, drawing on a contemporary case of platform -circulated synthetic imagery, that algorithmic systems of circulation, ranking, and amplification may reconfigure each dimension, allowing visibility to operate as a provisional source of authority and accelerating affective and legitimacy effects beyond human deliberation. Through analyses of iconic photographs and a case of AI -generated misinformation, the article compares historical configurations of visual politics with a contemporary, platform -mediated episode to clarify how operative logics shift under algorithmic visibility. The framework offers a transferable analytical vocabulary for examining political imagery in platform--mediated environments and across diverse cultural contexts.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[GenAI and Authorship in Academia: The (Im)Possible Creator. An Integrative Review]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2026-0004</link>
            <guid isPermaLink="false">/article/10.2478/doc-2026-0004</guid>
            <pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[Amid vivid ethical discussions on using AI to assist scholars in the writing process, this contribution tackles the question about Generative AI and authorship from a philosophy of language and academic perspective. In this integrative review, we look into the ways Generative AI impacts the very notion of authorship on the stages of knowledge production, knowledge consumption and knowledge dissemination. We focus on the implications of using Large Language Models with their enhanced capabilities of generating convincing imitations of human natural language for our conception of authorship. Contextualizing this capability to the realm of academic production, we seek to explore whether generative AI is going to further add to the postmodern destruction of the author, and how it relates to a core aspect of academic integrity, namely acknowledging the intellectual contributions of other scholars.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Algorithmic Kitsch: AI-GeneratedWritingas Aesthetic Simulation in Managerial Practice]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2026-0002</link>
            <guid isPermaLink="false">/article/10.2478/doc-2026-0002</guid>
            <pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[The proliferation of generative artificial intelligence (AI) in managerial practice has introduced a new paradigm of organisational writing that is fluent, coherent, and rhetorically effective, yet increasingly detached from human intentionality and lived experience. This article conceptualises this phenomenon as algorithmic kitsch, defined as AI -generated discourse that simulates meaning through aesthetic coherence and persuasive form while lacking epistemic depth and authentic subjectivity. Drawing on theories of kitsch from aesthetic philosophy and their extension into management studies, the paper situates AI -generated texts within a broader trajectory of organisational communication characterised by standardisation, emotional templating, and symbolic overproduction. It argues that generative AI does not merely assist managerial writing but intensifies pre--existing tendencies toward aesthetic simplification and rhetorical cliché, thereby transforming argumentation into a form of patterned plausibility. The article outlines the mechanisms of algorithmic kitsch (i.e., aesthetic smoothing, semantic flattening, and affective pre -structuring) and discusses their implications for epistemic integrity, ethical responsibility, and the erosion of authentic managerial voice. From a humanistic management perspective, the paper calls for greater consciousness in the use of AI -generated texts and renewed attention to aesthetic literacy and critical reflexivity in AI -mediated managerial communication practices.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[False Positive Risk in AI Detection of L2 Academic Writing: A Case Study]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2026-0003</link>
            <guid isPermaLink="false">/article/10.2478/doc-2026-0003</guid>
            <pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[AI -text detectors are increasingly used in education and scholarly publishing. While these tools are meant to protect academic integrity, they may also create new inequities for multilingual writers. This exploratory case study investigates why legitimate second language (L2) academic prose can be misclassified as AI -generated. We start from the premise that L2 academic writing functions as a distinct register whose recurrent linguistic features create predictability, which is one of the main metrics to detect AI -generated text. Using a single -author corpus repeatedly flagged as machine--generated, we develop a detector -inspired simulation capturing five proxies of high predictability: low perplexity, low burstiness, high formulaic language, consistent formality, and low lexical diversity. We simulated two common detection models: a prevalence model (flagging frequent proxies) and a cluster--based model (flagging co -occurrence of features). Prevalence -based detection flagged 62% of the sample manuscripts, while cluster -based detection flagged 38% with three or more co -occurring proxies. Although dense clustering (four to five proxies) was uncommon (7.8%), our sample was repeatedly flagged by both logics. These findings do not establish equivalence with commercial detection systems, but they suggest that highly conventionalized L2 academic prose may face a heightened risk of false positive AI flagging. We term this phenomenon the “detection paradox”.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Making Students “Safe” Writers with the SAFE Model in the Age of AI]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2026-0008</link>
            <guid isPermaLink="false">/article/10.2478/doc-2026-0008</guid>
            <pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[University students can develop opinion -writing skills through strong argumentative reasoning, while technological advancements should support high -quality text production without relying too heavily on AI text -generating tools. This study investigates students’ responses regarding the challenges and needs in writing argumentation -based opinion texts. Data were collected through a questionnaire administered to 97 students enrolled in the Indonesian Language Education Study Program in West Java, Indonesia. Data analysis employed percentage calculations based on the 6+1 Traits of Writing framework, encompassing ideas, voice, word choice, sentence fluency, conventions, and presentation. Argumentative reasoning was examined using the framework of claim, evidence, and reasoning. The findings reveal that students experience the greatest difficulty in writing conventions, while weaknesses in argumentative reasoning are primarily related to the presentation of evidence. To address these challenges, this study integrates the 6+1 Traits of Writing with an argumentation--based reasoning framework to develop the SAFE writing instruction model, consisting of Surveying, Arguing, Framing -Writing, and Editing. The SAFE model utilizes artificial intelligence as scaffolding in the writing process and responds to debates on text ownership in Generative Artificial Intelligence by positioning students as the primary agents of text production.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Algorithmic Storytelling and Vernacular Folklore: Rewriting the Nenek Pakande Monster in the Age of Generative AI]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2026-0010</link>
            <guid isPermaLink="false">/article/10.2478/doc-2026-0010</guid>
            <pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[The emergence of generative artificial intelligence has unsettled long -standing assumptions about authorship, originality, and authenticity in literary and cultural production. While debates often focus on academic writing or creative industries in the Global North, less attention has been given to how AI -mediated rewriting affects vernacular narrative traditions. This article examines the Indonesian folklore figure Nenek Pakande through the lens of ecological dramaturgy to explore how algorithmic text generation reshapes narrative voice and cultural memory. Drawing on conceptual analysis and comparative rewriting experiments, the study argues that AI does not merely redistribute authorship but transforms its ontological conditions by introducing non -experiential textual agency. In doing so, it reconfigures originality from inventive subjectivity to probabilistic recombination and reframes authenticity as a question of relational accountability rather than textual purity. By situating folkloric rewriting within the dynamics of algorithmic culture, this article proposes a model of hybrid ecological authorship that acknowledges technological mediation while maintaining ethical responsibility toward cultural and ecological specificity. Rather than positioning AI as either threat or tool, the paper advances a relational framework for understanding writing in the age of distributed cognition.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Working with ChatGPT: Engaging with the Real World and Our Own Academic Practice]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2026-0005</link>
            <guid isPermaLink="false">/article/10.2478/doc-2026-0005</guid>
            <pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[When using ChatGPT to write an ethnographic description of an intercultural event or to make a picture of a distant location, it by default employs the dominant discourse in intercultural communication of positivist, exoticist, Orientalist Othering. It also consciously uses modes of description which give the false impression that it has direct experience of social settings. However, when it is challenged, it displays impressively critical understanding of the ‘epistemic danger’ of such representations and why they are employed. This paper argues that this is therefore an example of AI representing the discoursal and ideological struggles which are already present in the academic field of intercultural communication and perhaps in qualitative social science more generally—where a dominant naïve positivism, encouraged by our neoliberal institutions, similarly, by default, takes us away from engaging directly with the social world we seek to investigate. We can also be reminded by this engagement with AI what it takes to deCentre how we engage with such realities— to apply a phenomenological interrogation of how we represent them. This is by no means a new task. AI simply provides a rapid example of its intense importance.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Authenticity and Authorship Beyond the Algorithm: Action, Institution, and Epistemic Responsibility under Generative AI]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2026-0006</link>
            <guid isPermaLink="false">/article/10.2478/doc-2026-0006</guid>
            <pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[This paper examines whether generative artificial intelligence can be considered an authentic author. While large language models (LLMs) demonstrate advanced linguistic competence and produce coherent academic discourse, they foreground unresolved questions concerning the ontology of authorship and the status of authenticity in scholarly authorship. The article is concerned with authorship rather than writing as such; writing is treated as the practical and institutional site through which authorship becomes visible, contested, and assigned.Rather than treating authorship as a property of textual output, the paper conceptualises it as an activity across three intersecting perspectives: praxis, institutional recognition, and epistemic responsibility. Extending Hannah Arendt’s The Human Condition, authorship is theorised in terms of linguistic labour, linguistic work, and linguistic action, with authenticity located in linguistic action as public, accountable, position -taking engagement. A Foucauldian perspective reframes authorship as an institutional function governed by discursive norms, while the epistemic perspective examines whether LLMs satisfy conditions such as semantic understanding, grounding, and epistemic commitment.The paper argues that AI can perform linguistic labour, support linguistic work, and approximate institutional markers of authorship, but cannot initiate linguistic action or assume epistemic responsibility.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[From Mental Concept to Algorithmic Exemplification: Rethinking the Aesthetic Situation in the Age of Generative AI]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2026-0009</link>
            <guid isPermaLink="false">/article/10.2478/doc-2026-0009</guid>
            <pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[The emergence of generative artificial intelligence (AI) necessitates a re -evaluation of the foundational structures of artistic creation and reception. This article utilises the “aesthetic situation” theory, originally formulated by Maria Gołaszewska (1984) and recently expanded through a managerial lens by Michał Szostak (2023b), to analyse the transformative impact of computational agents on the creative process. Building upon the distinction between “supporting” and “substituting” AI roles (Szostak &amp; Modliński, 2025), this paper introduces a conceptual model of “algorithmic exemplification”. It posits that generative AI functions as a technical mediator, translating an artist’s internal “mental concept” into a publicly accessible artefact through iterative prompting and latent -space navigation. By systematically analysing the five components of the aesthetic situation—the artist, the artwork, the recipient, the world of values, and the real world—this article argues that generative AI does not merely automate production but reconfigures the ontic status of the artwork and the agency of the creator. The discussion highlights the shift from direct craftsmanship to “aesthetic situation management”, in which the artist’s role evolves into that of a curator and prompt engineer, potentially substituting for traditional manual skills while supporting higher--order conceptual synthesis.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Rethinking AI in Writing Pedagogy: From Fluency to Rhetorical Competence]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2026-0007</link>
            <guid isPermaLink="false">/article/10.2478/doc-2026-0007</guid>
            <pubDate>Fri, 03 Jul 2026 00:00:00 GMT</pubDate>
            <description><![CDATA[Generative AI tools such as ChatGPT are reshaping academic writing in higher education, offering students unprecedented fluency, coherence, and accessibility. Yet these seductive characteristics, together with serious rhetorical limitations, raise concerns about authorship, integrity, and the development of critical thinking. The similarities and differences between AI generated texts and human writing, and the implications of these for writing instruction, however, remain underexplored. This study aims to bridge this gap by synthesizing five studies conducted by Jiang and Hyland, which systematically compare ChatGPT--generated essays with those authored by British undergraduates. The studies reveal a consistent pattern: while ChatGPT excels in producing structurally polished texts, it underperforms in rhetorical depth, interpersonal engagement, and authorial presence. Building on these findings, the paper proposes a framework for integrating AI into writing pedagogy. This positions ChatGPT as a supportive tutor rather than a proxy author, stressing both its advantages and limitations. It offers practical pathways for educators and underscores the importance of cultivating critical AI literacy, enabling students to navigate new writing contexts where human creativity and algorithmic fluency intersect.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Preface: The Linguistic Foundations of Management: Insights Across Contexts]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2025-0010</link>
            <guid isPermaLink="false">/article/10.2478/doc-2025-0010</guid>
            <pubDate>Tue, 23 Dec 2025 00:00:00 GMT</pubDate>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Why Words Matter: An Interview with Prof. Gail Fairhurst on Leadership and Discourse]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2025-0011</link>
            <guid isPermaLink="false">/article/10.2478/doc-2025-0011</guid>
            <pubDate>Tue, 23 Dec 2025 00:00:00 GMT</pubDate>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Language and Generative AI: A New Paradigm of Organizational Research]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2025-0012</link>
            <guid isPermaLink="false">/article/10.2478/doc-2025-0012</guid>
            <pubDate>Tue, 23 Dec 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[Language is not merely a medium of communication but a constitutive force in organization management. Three decades after the first “linguistic turn” in organization studies, generative artificial intelligence (GenAI) and large language models (LLMs) are provoking a second, data -intensive turn that reconfigures the relationship between language, technology, and management. LLMs now operate as discursive actors that simulate, generate, and transform organizational communication.This paper advances algorithmic discourse research as a new paradigm for studying language in organizations. It reframes metthodological rigor as pluralistic and reflexive, combining computational scale with interpretive depth. It retains traditional standards of evidence while extending them to encompass ethical and contextual reflexivity, acknowledging that meaning, data, and validity are co -constructed.An integrated multilevel framework links micro -linguistic forms (lexical, metaphorical, modal), meso -level routines and narratives, and macro -level outcomes such as innovation, trust, and performance. The new paradigm expands the methodological and epistemological foundations of organizational research by positioning language as both data and process, and LLMs as analytic partners in the study of sensemaking. In doing so, it marks a shift from observing discourse to co -engaging with algorithmic language, opening new avenues for understanding how organizations think, communicate, and act in the age of AI.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Through the Magic Lens: Management, Language, and AI]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2025-0013</link>
            <guid isPermaLink="false">/article/10.2478/doc-2025-0013</guid>
            <pubDate>Tue, 23 Dec 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[In this essay, we explore management, language, and artificial intelligence (AI). We highlight how management performs generative AI as magic and argue for the continued crucial role of humans under conditions of AI hype. We view language as an essential part of management to impress, to weave the illusion of expertise, and to manufacture our consent. We argue that performing AI as magic brings this into sharp relief, showing how power and ideology in and around AI operate when its advancement is constructed as inevitable. We encourage discussion on this and propose more research on resistance to management performing AI, on the gendered nature of AI and its consequences, and on AI -powered management clones and their language when agentic AI enters the stage. The magic lens offers a fruitful way forward in studying these fundamentally important topics.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[On Institutional Translation in a Fictional Context: Interdisciplinary Remarks on the Structure and Hierarchy of the Royal Institute of Translation in R. F. Kuang’s Babel]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2025-0022</link>
            <guid isPermaLink="false">/article/10.2478/doc-2025-0022</guid>
            <pubDate>Tue, 23 Dec 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[Employing an interdisciplinary approach from linguistics, literary studies, and translation studies, this article examines the structures and hierarchies present in the fictional Royal Institute of Translation, the emblematic institution of British imperial power which is the lynchpin of R. F. Kuang’s 2022 speculative novel Babel. With the work primarily set in a fictional 1830s Oxford where magical translation -based silver bars are the key to colonial dominance, the institutional management of languages and translation unsurprisingly assumes a key role. After contextualising Kuang’s towering creation in historical, linguistic, and literary terms, the hierarchy and structure of the Institute is presented and discussed via close reading of a selected excerpt from the novel. This analysis is complemented by relevant observations on the institutional and professional context of translation and interpreting both in historical and in contemporary times.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Crisis Leadership: Language, Power, and the Construction of Legitimacy in War Narratives]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2025-0016</link>
            <guid isPermaLink="false">/article/10.2478/doc-2025-0016</guid>
            <pubDate>Tue, 23 Dec 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[Focusing on wartime narratives, this study demonstrates how language functions as a tool of power and legitimacy, revealing discursive patterns that operate not only in politics but also in corporate and institutional contexts.Using a qualitative CDA approach, the paper analyses how political leaders employ language during crises to justify decisions and manage military operations. Through comparative analysis of recent scholarship and our examination of speeches and statements by top political figures involved in the wars in Ukraine and Gaza, the study identifies conflict framings, recurring themes, linguistic choices, and rhetorical devices used to legitimise “Us” and delegitimise “Them”. The findings indicate that crisis leaders rely heavily on the victim—savior dichotomy, constructing “Us” through themes of victimhood and moral superiority, and depicting “Them” through savagery, and moral decline—compounded by marked evaluative lexis, specific syntactic patterns, and frequent rhetorical devices.The paper argues that understanding how leaders construct wartime narratives to justify violence can enhance awareness of analogous discursive strategies in organisational crisis contexts and support more ethical, transparent, and responsible communication practices.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Beyond the Buzzwords: Why Ideas and Discourse Matter in Management]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2025-0017</link>
            <guid isPermaLink="false">/article/10.2478/doc-2025-0017</guid>
            <pubDate>Tue, 23 Dec 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[The social sciences have seen significant changes over the past three decades, with a return of ideas and discourses as crucial factors in management and organisational studies. This article examines the critical frameworks and analytical strategies for studying language and power that have been applied to the management field. We approach this by revisiting the literature on public administration, the domain where discourse first emerged as a key variable in the social sciences. Our contribution is an extensive review of the international literature to quantitatively and qualitatively assess academic output in this area. Based on the “argumentative turn”, we argue for three socially mediated dimensions—logistic, ideational, and material—that act as an interdisciplinary link with linguistics. We argue that discursive approaches are essential to management because they anchor the field in constructivism, thereby highlighting the power of ideas in the analysis of institutional dynamics.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Guardians of Grammar, Gatekeepers of Gender: A Critical Discourse Analysis of the Académie Française’s Report on the Feminization of Professional Titles]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2025-0020</link>
            <guid isPermaLink="false">/article/10.2478/doc-2025-0020</guid>
            <pubDate>Tue, 23 Dec 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[This article examines how the Académie française constructs its institutional position on the feminization of professional and official titles in French. Using Fairclough’s three-dimensional model of Critical Discourse Analysis (CDA) and integrating Feminist CDA (Lazar, 2005), the study analyzes the Académie’s 2019 report as a case of symbolic discourse management. It investigates how feminization is framed linguistically, discursively, and ideologically. At the textual level, the report employs hedging, modality, and passive constructions to convey caution and deflect institutional responsibility. At the discursive level, the Académie constructs an image of neutrality and methodological restraint while subtly delegitimizing feminist discourse through rhetorical distancing. At the sociocultural level, feminization is normalized in lower -prestige professions but resisted in elite or masculine -coded domains, maintaining symbolic gender hierarchies. Legal and international considerations are mobilized to justify procedural caution. The analysis reveals three cross--level patterns: cautious distancing, symbolic boundary maintenance, and discursive deferral. These findings demonstrate that the Académie’s discourse not only reflects but actively manages institutional authority, gender norms, and the boundaries of acceptable linguistic change. The study contributes to debates on language policy and gender equality by framing feminization as a site of ideological governance rather than grammatical evolution.]]></description>
            <category>ARTICLE</category>
        </item>
        <item>
            <title><![CDATA[Generative AI and Generation Z: Redefining Language, Identity, and Communication in the Digital Workplace]]></title>
            <link>https://reference-global.com/article/10.2478/doc-2025-0014</link>
            <guid isPermaLink="false">/article/10.2478/doc-2025-0014</guid>
            <pubDate>Tue, 23 Dec 2025 00:00:00 GMT</pubDate>
            <description><![CDATA[This paper examines the intersection of Generation Z, artificial intelligence (AI), and workplace communication, focusing on how digitalization is reshaping professional discourse, leadership, and identity. As the first fully digital -native cohort, Generation Z operates within hybrid work environments increasingly mediated by AI -driven tools such as ChatGPT and Microsoft Copilot. These technologies influence tone, feedback, and collaboration, enhancing efficiency while simultaneously challenging authenticity, empathy, and intergenerational understanding. Gen Z’s communicative repertoire—marked by multimodality, informality, and visual semiotics such as emojis, memes, and abbreviations—serves both identity construction and peer bonding. Through a narrative review, the paper explores work values, collaboration preferences, and adaptive communication styles, emphasizing the importance of aligning communicative norms with the expectations of emerging generations. As a conclusion, it argues that AI functions not merely as a productivity enhancer but as a transformative linguistic force, demanding that workplaces balance technological fluency with inclusivity, relational connection, and ethical communication practices.]]></description>
            <category>ARTICLE</category>
        </item>
    </channel>
</rss>