AI's Blind Spot: Language Worldviews to the Rescue

AI's Blind Spot: Language Worldviews to the Rescue | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Systematic Review AI's Blind Spot: Language Worldviews to the Rescue Tetiana Ilman This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6719644/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract This paper examines the intersection between artificial intelligence (AI), linguistic diversity, and cultural cognition through the lens of linguistic relativity. It argues that language is not merely a communication tool but a fundamental framework shaping human cognitive perception, memory, and social behavior. Using a contrastive analysis of word formation in English and Ukrainian, the study illustrates how different linguistic structures reflect distinct worldviews. It further examines how AI development, centered mostly around English and other dominant languages, risks reinforcing cultural and linguistic homogenization. Experimental testing demonstrates how AI-generated outputs fail to capture culturally embedded emotional and metaphorical distinctions, particularly in underrepresented languages. The paper also highlights that the loss of a language entails the loss of a unique worldview and a diminished capacity for cognitive comparison. Linguistic and cultural diversity enable self-reflection, enhance communication, reduce bias, and strengthen social justice and identity. In response, the paper proposes the creation of an Atlas of Language Worldviews - an AI-enhanced platform to systematically document, preserve, and map cultural perspectives across languages as well as to train AI. The Atlas would anchor a “worldview” with measurable components and integrate linguistic, historical, and cultural data, which would offer a critical tool for supporting cultural self-reflection of language groups, intercultural understanding, preserving endangered cultures, and training the ethical development of culturally sensitive AI systems. The paper provides the roadmap for Atlas creation. Artificial Intelligence and Machine Learning Linguistic relativity Language Worldview Cultural diversity Artificial intelligence Compound word formation AI homogenization Language Worldview Atlas 1. Introduction My fascination with the link between language and culture began at the age of six, with a Russian book The Tales of the Peoples of the World. I was struck by how vastly the stories differed from familiar Russian and Ukrainian ones - especially those from Australia, Oceania, eastern tribes. Names, humor, life goals, and values - all felt alien yet captivating. For me, that foreignness manifested in the strange names they gave to people, objects and phenomena, in the way they spoke. Even then, I could sense the connection between the language and their worldview. This early intuition evolved into an academic pursuit. I began to explore an ancient yet unresolved question: do cultural/environmental differences cause languages and worldviews to differ, or vice versa: maybe the language is some type of encoded frame that shapes human cognition and worldview? This question has fascinated thinkers for centuries, from Plato, St. Augustine, Humboldt, Boas, Sapir, Whorf, Wittgenstein, Vygotsky to modern cognitive linguists (Lakoff, Boroditski, Turbayne), specialists in language ideology, language behaviour, etc. Having reconsidered early ideas of linguistic relativity, contemporary studies across psycholinguistics, cognitive science, and anthropology continue to demonstrate that language reveals deep cognitive and cultural differences - what we may call language-specific worldviews. It influences how we perceive time, space, emotion, and social relationships and build our behavior and knowledge categorization. In my work I provide impressive worldview diversity examples from contemporary linguistic relativity studies as well as conduct my own research in how language structures (on the example of English and Ukrainian suffix derivatives and compounds) reveal embedded differences in cultural cognition. Derivation and compound word formation was not a randomly chosen sphere of contrastive worldview analysis. In human speech, naming activity is the most telling aspect from the point of view of the correlation of language units with the extralinguistic objective world. Naming in language is always meaningful . As languages of the Indo-European family, English and Ukrainian are both productive in forming compound words. Therefore, the c ontrastive analysis of English and Ukrainian worldviews forms the first line of research . The idea of the second line of research was motivated by the paradoxical situation we are facing. On one hand, artificial intelligence - especially large language models (LLMs) - has revolutionized access to knowledge, accelerating learning and linguistic processing at an unprecedented scale. On the other hand, these technologies pose a serious threat to linguistic and cultural diversity, which causes AI homogenization . AI systems, often trained primarily on English data, are not culturally neutral. They encode and reproduce dominant cultural perspectives, flattening nuanced worldviews into standardized outputs. Many languages have not yet crossed the digital divide, and may never do so unless conscious steps are taken. We can state that we live in an era where we have not yet fully understood the worldviews encoded in languages, and at the same time we are already at risk of losing those nuances, which means we are losing a crucial tool of self-reflection. In our world cognition advances through comparison. Without contrasting categories from different cultures and languages, human understanding becomes narrower and less reflective. Comparing the culture of the language group to other cultures is an efficient tool for self-reflection. When cultures examine themselves in the mirror of another language, they not only refine their self-awareness, but also reduce bias, improve communication, strengthen identity and better preserve valuable traditions. Without such reflection, cultural assumptions remain unchallenged and unexamined. The loss of a culture or a language is seen as not merely the disappearance of another worldview system but rather as the loss of a unique lens of world perception. To illustrate this problem, a contrastive analysis of English and Ukrainian worldviews is held along with the second line of research : an experimental investigation into how multilingual AI tools reflect (or fail to reflect) these linguistic worldviews in generated content. The findings reveal a critical blind spot in AI development - its insensitivity to the cultural and cognitive dimensions encoded in language .The study explores how AI’s reliance on English-centered data and its preference for efficiency over nuance flattens cultural distinctions. As a solution, I suggest the development of a Language Worldview Atlas - a multilingual platform for documenting, comparing, preserving cultural perspectives across languages, as well as for AI training. This project aims not only to address worldview erasure but to create a foundation for ethical and culturally aware AI systems. 2. Theoretical Background Linguistic relativity The idea that language shapes thought has a long intellectual history, touching philosophy, anthropology, and modern cognitive science. Early thinkers such as Plato, St. Augustine, and Roger Bacon considered language to be central to understanding reality. In the 18th and 19th centuries, Wilhelm von Humboldt and Johann Gottfried Herder proposed that language expressed the "spirit" of a people - the insight would later influence the theory of linguistic relativity. In the early 20th century, this line of thought gained momentum through the work of Franz Boas and his students, notably Edward Sapir and Benjamin Lee Whorf. Whorf studied the language of Hopi (one of the Native American tribes) and suggested it had a different concept of time than “Standard Average European” [ 25 ] speakers had. He gave arguments that Hopi had "no words, grammatical forms, construction or expressions that refer directly to what we call 'time' and concluded that the Hopi had "no general notion or intuition of time as a smooth flowing continuum in which everything in the universe proceeds at equal rate, out of a future, through the present, into a past." [24] These particular ideas for time category in Hopi were later refuted [16]. However, the theory of linguistic relativity, often referred to as the Sapir-Whorf hypothesis, has modified and become a reliable reference for many interdisciplinary studies, like psycholinguistics, cognitive linguistics, behavioral linguistics. Contrastive psycholinguistic studies have shown that speakers of different languages conceptualize time, space, and color in ways aligned with the structure of their native tongue, for example, that timelines come in all shapes and directions. For the Aymara people of the Andes, time flows front to back. The past, which was known and hence seen, lies in front. The unknown and unseen future is behind [21]. Mandarin speakers (China) sometimes represent time along a vertical axis, the past is above and the future is below [9]. According to new studies in language ideology [5], in the field of linguistic anthropology cultures can be divided into two groups based on the way that members of that culture generally perceive time: monochronic cultures and polychronic cultures. Cultures, called monochronic, tend to believe that time is linear. They like to do one thing at a time, stick to schedules, and value being on time (as in the U.S., Germany, and Great Britain). Other cultures, called polychronic, tend to see time as fluid and malleable, they are more flexible with time, don't mind interruptions, and prioritize relationships over strict schedules (as in Mexico or Egypt). Monochronic cultures value efficiency, while polychronic cultures are more adaptable and focus on people. In monochronic cultures these behaviors are thought to be inefficient and improper. People of polychronic cultures are more susceptible to distractions and open to interruption but are better at focusing on many tasks at once. Wittgenstein’s famous dictum -“The limits of my language mean the limits of my world”[1] - means that the scope of our understanding and experience of the world is fundamentally shaped and constrained by the language we use to describe and think about it, it captures the philosophical essence of linguistic relativity. If a concept cannot be named in one’s language, it becomes harder to notice, describe, or even think about. This is not merely a limitation of vocabulary but of conceptual worldview. Empirical evidence supports this notion. A well-known 2007 study on color discrimination revealed that Russian speakers (whose language make obligatory distinction between light blue -"goluboy"- and dark blue -siniy") were not only faster at distinguishing between these shades than English speakers (whose language uses the single word “blue” for both), but rather that Russian speakers couldn’t avoid distinguishing them [12] . This communicative requirement appears to cause Russian speakers to habitually make use of this distinction even when performing a perceptual task that does not require language.The study suggested that linguistic representations normally interfere in objective perceptual decisions. Other striking examples come from the language of the Kuuk Thaayorre people in Australia. They use cardinal directions (north, south, east, west) instead of “left” and “right”. It shows that their spatial perception is rather Earth-oriented, unlike ours - self-oriented. A speaker might say, “There’s an ant on your southwest leg”.This linguistic habit cultivates exceptional spatial orientation skills. "Hello" in Kuuk would literally mean, "Which way are you going?". The answer could be, "North-northeast in the far distance. How about you?" This constant need to stay oriented allows speakers of this language to maintain a strong sense of direction, even without special gadgets, like compasses, demonstrating that humans are capable of much better orientation than previously thought if trained by their language and culture. There is a distinctive peculiarity in how the speakers of Kuuk Thaayorre Aboriginal group organize time based on cardinal directions. When asked to arrange pictures of a person aging, they do not order the sequence from left to right or right to left (as English or Hebrew speakers might). Instead, they orient the timeline based on the direction they face, suggesting that their sense of time is landscape-centered rather than self-centered. Indeed, why would we organize time self-centered if we remember that the sun rises on the east and goes down on the west? Must be so selfish from Kuuk’s point of view. The Holy Roman Emperor, Charlemagne, said, “To have a second language is to have a second soul”, this is a strong statement supporting the idea that language crafts reality. Similarly, “as many languages you speak, as many times you are human/ the more languages you speak, the more times you are human”. The sayings are attributed to different people - Chekhov, Goethe, Masaryk. All these sayings mean that learning a new language is about entering a new culture, understanding different ways of thinking, and seeing the world through another lens. 2.1.From Chomsky’s Universal Grammar to AI “imperialism” The rise of cognitive science in the latter half of the 20th century has shifted the focus of studies from semantics to syntax, to generative approach. Noam Chomsky pioneered the hypothesis that syntax was the level reflecting a structure of the human mind. Chomsky's Universal Grammar (or General Grammar) theory developed primarily by Chomsky [10] suggested that every person has an innate mental structure of production and understanding of language, a set of abstract principles and parameters that underlie all human languages, or "universal grammar". The goal was to make a complete model of the inner language, and the model could then be used to describe all human language and to predict if any utterance would sound correct to a native speaker of the language. The Universal Grammar theory has provided the instrumental for AI creation, particularly in the areas of natural language processing (NLP) and understanding, as well as machine learning algorithms, which are a critical part of many AI systems. These algorithms are designed to recognize patterns and make predictions based on data. It provided a foundation for the development of AI systems that can process, understand natural language and respond to human language in a more sophisticated, natural, intuitive and human-like way. AI appeared in 2023 and its development has been explosive. It has been changing the world since then. It has turned into a revolutionizing instrument used at every level of human life. AI-centered scientific investments prevail considerably at this time, and this tendency will most likely only accelerate. World AI leading countries take part in the race for AGI ( Artificial general intelligence , which is a type of AI that matches or surpasses human cognitive capabilities across a wide range of cognitive tasks). Science these days is trying to figure out how “homogenezative” AI can be for culture and creativity, and language diversity [17]-[19]. AI has the potential to diminish culturological differences between languages and creativity: The main reasons for this to happen can be identified as: ● Coding Languages and Terminology : ○ English is the Dominant Language : English has been the dominant language on the internet since its beginning, largely due to its widespread use in business, education, and technology. English has been Lingua Franca (dominating) geographically and across domains. Almost everything we know now about the human mind and human brain is based on studies of usually English-speaking “undergraduates at universities” [7]. Much of the codebase, documentation, and technical literature in the AI and technology world is in English. Even as AI is used to translate and localize these resources, it still subtly reinforces the dominance of English in these domains. Users of AI, developers or end-users, are frequently exposed to English terminology, subtly shaping the way they think about technology and its potential. This can impact how other languages are used to describe similar concepts, potentially borrowing English terms rather than developing native ones. ● Translation/ Interpretation : ○ Dominance of Certain Languages : Current translation AI models are often trained more extensively on English, major languages, like Mandarin, and Spanish. This can lead to translations into these languages being of higher quality and easier to understand. Over time, individuals may be subtly predisposed to communicate primarily in these dominant languages to ensure their message isn't lost in translation, further marginalizing less common languages and their associated cultures. ○ Homogenization of Meaning : AI-powered translation tools, even though it has been constantly improving, can sometimes flatten nuances, idioms, and cultural references that are deeply embedded in a language. They might prioritize conveying the core meaning but miss the subtle undertones, emotional weight, or historical context that make a phrase or expression unique in its original language. Over time, reliance on these tools could lead to a standardization of language, where complex culturally-specific terms are replaced with more universally understandable alternatives. ● Content Creation and Consumption : ○ AI-Generated Content in Other Languages : AI can generate content (text, images, video, etc.) in various languages. However when the AI is trained primarily on a single cultural dataset (likely a Western dataset), it might implement biases and cultural assumptions into the content it produces, even when generating in a language associated with a different culture. This is a form of cultural “imperialism” via AI [6]. This could lead to a convergence toward a more globalized, homogenized cultural aesthetic and value system. ○ Personalized Recommendation Systems : AI systems already that recommend books, movies, and music from around the world. While it can be a great tool for cultural exchange, if these systems are designed to enhance engagement, they push users toward content that aligns with their pre-existing preferences (often shaped by their local culture). This could reduce exposure to diverse cultural viewpoints and expressions, reinforcing existing echo chambers and inhibiting true cross-cultural understanding. The Forbes article "AI Homogenization Is Shaping The World" by Hamilton Mann (Forbes, 2024) [17] argues that the increasing reliance on a small number of large AI models, particularly from major tech companies, is already leading to a homogenization of thought and creativity. ● Language Learning : ○ Standardized Language Instruction : AI-powered language learning tools can be incredibly effective, but they can also promote a standardized version of a language. They may focus on grammar and vocabulary that are considered "correct" or "common" potentially neglecting regional dialects, slang, and other forms of linguistic diversity that are essential facets of a culture. ○ Loss of Cultural Immersion : While AI-tutoring can enhance learning, it cannot fully replicate the experience of cultural immersion that is crucial for understanding the subtle nuances and context behind a language's usage. Learning a language is more than memorizing words; it's about understanding the cultural context from where it springs. The sad news is that linguistic studies show that, of the approximately 7,000 languages spoken today, about 2,500 are generally considered endangered. “Less than 5% of all languages can still ascend to the digital realm” [14]. Digital diglossia puts low-technological languages at risk of extinction. Many languages haven’t yet crossed the digital divide for many reasons. Some of them have little or no Internet access; some of them do not have a well-developed research community, some don’t get support from public administrations, etc. Young speakers of minor languages are growing up in conditions with the Internet and a global world with languages like English, Spanish, Chinese and other major languages being pervasive for virtual communication, digital and social media. Didactic materials in major languages are effortlessly available over and above those of native minor languages. The risks have been paid attention to. Since 2019 Unesco Headquarters in Paris have been holding conferences devoted to Language technologies (Language Technologies for All - LT4All 2019). The 1st one, themed “Multilingualism for Building Knowledge Societies”, highlighted the critical role of language and cutting-edge technology, including artificial intelligence, in shaping cross-cultural communication. The conference resulted in certain initiatives by research institutions and technological companies toward developing language technologies for a wider range of languages. “Despite significant progress, however, many communities are still being left behind” [23]. The 2nd International Conference (LT4All 2025) themed “Advancing Humanism through Language Technologies” and aimed at furthering the agenda of language technologies with a focus on community empowerment [19]. At the LT4All Conference in March 2025, it has been reported that even though: ● Large multilingual datasets are being created (like The OSCAR project - Open Super-large Crawled Aggregated coRpus) ● LLMs keep getting larger and (some of them) more diverse ● Meta developed the first multilingual translation models that don’t rely on English as pivot (2020-2022) ● Generative AI can create content in different languages from the start, dispensing with the traditional translation and localization process. STILL, HOWEVER the future of language (and cultural) diversity in the age of AI is uncertain and the challenges should be addressed with proper awareness and action on different levels: the language and the research communities, governments and public institutions and all of us. 3. Methodology 3.1 Linguistic Contrastive Analysis This study draws on an extensive contrastive analysis of word formation in English and Ukrainian. The research approach is grounded in the assumption that naming practices - especially compounding - reflect how speakers of different languages conceptualize the world. Naming is always mediated by thought. A compound word, as a unique unit of nomination, primarily indicates that it is enough for a speaker to name two related objects for the listener to independently guess the type of connection the speaker had in mind. Understanding the patterns of nomination leads to a deeper understanding of the role of the human factor in language. Any complex word can be considered a frozen context for its elements, therefore, analyzing those compound words in English and Ukrainian promise to reflect national specificities in perception and understanding of the world. A new derivative, being a creation of word-formation act, like any new word, represents the product of the generalizing mental activity of a person, a product of understanding the features of the object being named, a result of singling out one feature from many, and thus a result of abstraction. After all, it is undoubtedly that the derivation process and compounding (these processes that can be directly observed in living speech) do reflect, as a mirror, numerous characteristic features of the act of nomination. Thus, it's generally believed that most, if not all, natural languages have some form of compounding. Compounding is observed across different language families and linguistic typologies. Particular attention in the work was devoted to observing the semantic connections between objects of reality in the process of word-formative nomination in English and Ukrainian and, with its help, there was an attempt to compare two linguistic "worldviews" (understandably, fragmentary and illustrative) - English and Ukrainian. The contrastive method in the research was used in combination with the technique of decomposition or reconstruction of the word-formation act . Both languages studied, English and Ukrainian, form compounds using similar word classes (nouns, adjectives, verbs, adverbs). As languages of the Indo-European family, they are both productive in forming compound words. The analysis was based on data drawn from multiple sources, including general and specialized dictionaries, literary texts, journalistic materials, and examples from everyday speech. Emphasis was placed on compounds that are lexicographically fixed, emotionally marked, or culturally embedded. The analysis focused on three semantic fields: Nature and Environment Vocabulary Compound names related to flora, fauna, geography, and the domestic environment. These terms often reveal the associative patterns speakers use to connect physical traits with metaphorical or functional significance. Anthropocentric Vocabulary Compounds that describe human attributes, behaviors, or roles. These often rely on metaphor, tactile imagery, and evaluative connotations. Attention was given to how attributes like strength, intelligence, or emotionality are encoded differently across languages. Culturological Vocabulary (Culturemes) Words that reflect socially specific concepts, practices, or values. These include idiomatic formations, culturally loaded metaphors, and traditional expressions that encapsulate the worldview of a linguistic community. Ukrainian, a synthetic language, relies more heavily on suffixation for emotional and diminutive expression, while English, an analytic language, leans on compounding for pragmatic categorization. The comparative study examined not only the structure of compounds but also the emotional, metaphorical, and evaluative layers embedded in them. Special attention was paid to diminutive-hypocoristic forms, imperative holophrases (especially in surnames and nicknames), and culturally resonant occasionalisms (creative neologisms). The goal was to reveal not just linguistic differences, but underlying cultural tendencies. 3.2 Multilingual AI Output Evaluation To complement the linguistic analysis, a second research line focused on testing how AI systems reproduce culturally embedded distinctions in compound word formation across languages. The objective was to determine whether large language models trained predominantly on English could effectively capture and convey the culturally nuanced cognitive strategies embedded in other languages - particularly Ukrainian, Russian, and Polish. An experimental setup was created using publicly accessible AI tools that simulate multilingual capabilities. The same prompt - ‘culturological differences in English and Ukrainian compound word formation’ was submitted in four languages: English, Ukrainian, Russian, and Polish. The goal was to observe what kind of information the AI provided in each language and whether the outputs mirrored the contrastive tendencies identified in human-generated linguistic data. The prompt was intentionally framed in a high-level analytical style to encourage AI to generate content beyond simple definitions or grammatical explanations. The outputs were then qualitatively analyzed for thematic depth, emotional and metaphorical framing, structural complexity, and cultural relevance. Key observations included: Russian AI Output : Reflected a nuanced understanding of linguistic worldview concepts. It captured emotional tone, cultural embeddedness, and metaphorical layering similar to the author’s own findings. This was likely due to greater availability of training data on this topic in Russian academic and cultural domains. Ukrainian and Polish Outputs : These were largely similar to each other, often mirroring each other's phrasing and structural content. While they acknowledged emotional and metaphorical differences between languages, they tended to reduce Ukrainian linguistic creativity to folk or rural imagery and emphasized agriculture-related vocabulary. The outputs appeared limited by a lack of rich training data and a tendency to replicate formal or encyclopedic language. English AI Output : Focused on technical parameters such as structural types, productivity, joining elements, and frequency. It introduced novel categories like “tolerance for linguistic innovation” and “influence of ideology and politics” but failed to convey emotional or metaphorical subtleties. The result was a pragmatic, stylistically neutral treatment of a topic that is inherently culturally charged. The comparative AI outputs revealed a consistent pattern: emotional depth, metaphor, and cultural humor were most likely to be omitted in English outputs, whereas Slavic-language outputs - especially in Russian - reflected richer semantic and cultural nuance. These differences underline a core limitation of current multilingual AI systems: while they may translate structure or meaning, they often fail to preserve the cultural worldview encoded in linguistic form. This finding supports the broader hypothesis that AI, when not trained on culturally diverse datasets, reflects the biases and limitations of the dominant language paradigms it is built upon. It raises critical questions about the ethics of AI language modeling, especially as such tools increasingly shape how knowledge is generated, translated, and consumed worldwide. 4. Findings 4.1 English vs. Ukrainian Compound Structures The contrastive analysis of word formation (suffixation and compounding) in English and Ukrainian reveals fundamental differences in how each language encodes perception, emotion, and social values. These differences are not merely linguistic but cultural and cognitive, illustrating distinct worldviews. Ukrainian (as a synthetic language) turned out to be significantly less saturated with compound new words than English (as an analytic language), because suffixation is paramount in its derivational system. Emotional and Diminutive Derivation The most striking divergences in word formation in English and Ukrainian lie in the emotional expressiveness. Ukrainian, like other Slavic languages, makes extensive use of diminutive-hypocoristic forms. These are formed not only from nouns but also from adjectives, verbs, and adverbs, using a variety of suffixes to express affection, tenderness, irony, or sarcasm. Examples: Іваночко (dear Ivan), червоненький (reddish with affection), швиденько (quickly, softly), and їстоньки (to eat, tenderly). English, by contrast, has a far more limited emotional derivational system, typically confined to select noun forms (e.g., doggie , duckling , darling ) and lacking morphological tools to create emotional forms from verbs or adjectives. Nature and Environment Vocabulary As for the semantic groups of compounds distinguished, floral compound words (from the first semantic group) turned out to reflect the most in common as for associative connections between the definitum and definition for English and Ukrainian word-making. Among the reasons we can outline is the fact that a significant number of names for elements of the surrounding world were inherited from classical languages. Another reason for that was the common European environment and easy migration for cultures and languages on that territory, which caused similar associations to be evoked in representatives of two linguistic communities. For example: crane’s-bill – “журавець” (“zhuravets”, literally "crane" +suffix), cotton-grass – “пухівка” (“pukhivka”, literally "fluff +suffix"), dog-berry - “вовчі ягоди” (“vovchi yagody”, literally "wolf berries"),finger-flower - “наперстянка” (“naperstianka”, literally "thimble flower"), etc. Similarities in associations underlying the formation of names were also found in animal names. Anthropocentric Compounds Analysis of attributive composites (with attributive relationships between components) showed that English words are characterized by a significant prevalence of tactile associations. Most English composites denoting tactile associations with named objects correspond to Ukrainian composites or phrases with attributes indicating size, shape, color, aesthetic evaluation, for example, words with the following first components: hard -: -faced - похмурий, -featured “з різкими рисами”, -fisted - скупий, - grained - крупнозернистий, -nosed “тверезий, практичний”; heavy -: -browed “1. з похмурим обличчям, 2. з густими навислими бровами”, -handed “незграбний, невправний”, -headed “1. сонний, в’ялий, 2.похмурий ; light -: -fingеred “злодійкуватий” (it is interesting, that these equivalent have different shades in connotations - positive in English, and negative in Ukrainian), -footed “прудкий, швидконогий”, -handed “спритний, умілий”, -heeled “швидконогий”; crack -headed (-brainеd) “недоумкуватий”, broken -bellied - “що хворіє на грижу”, etc. Special attention in the study was devoted to English compound words with the components - heart - and - mind -, which in Ukrainian words or phrases are corresponded to by formations with the lexical unit - душ - (- dush - literally “ soul ”) and its synonyms, for example, words with the element -heart-: heart-strings - “тайники душі”, black-hearted – “з чорною душею”, heart-blood - “1. життєва енергія, 2. найдорожче”, large-hearted - “великодушний”, marble-hearted, iron-hearted - “бездушний”, down-hearted - “занепалий духом”, open-hearted - “1) щиросердечний, 2) з відкритою душею”, plane-hearted - “простодушний”, heartsease (heart’s-ease) – “душевний спокій”, heart-sick – “занепалий духом”, heart-struck – “зворушений до глибини душі”, false-hearted – “двоєдушний”, simple-hearted – “простодушний”, single-hearted – “прямодушний”, hen-hearted – “легкодухий” тощо, With - mind - element: noble-minded – “великодушний”, pure-minded – “чистий душею (серцем)”, single-minded – “прямодушний”. Both Ukrainian and English reflect a common understanding of the central role of the heart as a place of focus for emotion and moral qualities. However, as we can see, in the Ukrainian linguistic picture, the element - душ - (-soul-) plays a more important role than the heart, reflecting the enduring belief of Ukrainians in the existence of the soul as an immaterial component of a person. This suggests that English leans toward observable traits, while Ukrainian invokes internal states and moral essence. Culturological Vocabulary (Culturemes) Analysis of the nominative function of occasionalisms in Ukrainian and English indicated a predominantly anthropocentric nature of occasional nomination in both languages - English and Ukrainian. The most representative thematic groups of neologisms were found to be: "name of a person" (26.2% in English and 31.6% in Ukrainian). The most relevant function of the occasional word is evaluative (which are approximately 75% of English units of this class and 81% of Ukrainian ones), with the predominance of negative evaluations over positive ones being obvious (in a proportion of approximately 4:1 for both languages) Even now, in the age of Internet and AI homogenization, the new compounds that appear in Ukrainian and English show huge culturological differences. Many Ukrainian occasionalisms/neologistsm are related to war reality. Here are some of the most glaring examples: рашизм (rashism: Russian + fascism), Путлєрнет (Putin + Hitler + Internet), and Затридні (“in three days,” mocking the unrealistic timeframe of Russia’s initial invasion claims) show the language’s capacity for expressive, politically charged compounding. These creations reveal not only linguistic ingenuity but collective emotional and cultural resistance. English, by contrast, shows a generation of compounds tied to digital life and consumerism - doomscrolling , clickbait , blogosphere - reflecting a cultural focus on information saturation and technological immersion. Imperative Holophrastic Compounds Even though Ukrainian turned out to be much less productive in compounding in general, the observation yielded interesting results for one group of compounds that turned out to be more productive in Ukrainian than in English. Those are imperative holophrasis, representing instructions for the functional use of the named object or phenomenon: for example, wash-and-wear, cash-and-carry , pack and carry, keepsafe, gotomeeting, do-it-yourself, pay-as-you-go, pickme, etc. Holophrases of an imperative nature are inherently characteristic of the Ukrainian language, as evidenced by Ukrainian surnames and nicknames. They are usually built on the following model: a verb in the imperative mood followed by a noun in the nominative case, which is usually the direct object of that verb. The combination of these two elements, often unexpected and sometimes even absurd, is a testament to the unique Ukrainian humor. Just see the examples of some nicknames: Дунь-плюнь (“Dun’-plyun’ - lit. “blow and spit”, the name for someone who practices witchcraft), Ломинога (lomynoga - lit. “break leg”,the nickname for a daring dancer or a strongman who was capable of knocking down any opponent with a single blow, also for someone who was clumsy and awkward), дядько-дістань-горобчика (diad’ko-dostan’-horobchyka - lit.“uncle-get-a-sparrow”, for someone who is very tall, дівчина-розплітай-коса (divchyna-rozplitai-kosa -lit. “girl-unweave-the-braid”, for a girl of loose morals as braid was the symbol of virginity), Петро-бережи-ніс ( Petro-berezhy-nis - lit. “Peter-watch-your-nose,for someone who was clumsy and awkward). The last names that originated from the nicknames: Заплюйсвчка (Zapluisvichka - lit. “spit-candle”, for a buzz-kill),Засядьвовк (Zasiad’vowk - lit. “ambush-wolf”), Перебийнис (Perebijnic - lit. “break-nose”),Нагнибiда (Nagnybida - lit. “bend down-misfortune”), Неїжборщ (Nejizhborsch - lit. “don’t eat borsch”),Hепийпиво (Nepyjpuvo - lit. “don’t drink beer”),Неїжхлiб (Nejizhkhlib - lit. “don’t eat bread”), etc. These types of compounds were typical characteristics for Old Ukrainian, unlike Old English. The earliest composites of the type pickpocket (1300) are said to arise in the Middle English period under the influence of French imperative words (where they also appeared primarily in anthroponymy), characteristic of the colloquial style of speech [3]. Unlike nicknames and last names reflecting Ukrainian ironic emotional attitude to characteristics people named, Ukrainian compound first names reveal magic-like attempt to program the future behaviour of the baby, and even their destiny, eg. Miroslav is derived from Pre-Slavic and is typically given to boys. Miroslav is derived from the words “mir” (lit.‘peace’ or ‘world’) and “slava” - (lit. “glory, fame”), altogether it means “the glory for the world”, or “the glory of peace”. Волеслав (Voleslav -lit. “ glory of reedom”) , Доброслав (Dobroslav - lit. “glory of the good”), Осмомисл (Osmomysl - lit. “the one who has 8 thoughts, wise”, Броніслав/Боронислав (Bronislav/Boronislav -lit. “glory of defense), Владислав (Vladyslav / Volodyslav, meaning "glory of power" ,Володимир (Volodymyr - meaning “owning the world”, or “ruling in peace”. The comprehensive approach of Ukrainian and English compound word making analysis that was performed resulted the following conclusions: ● Ukrainian compounds possess greater imagery in comparison with the English language, which is related to the more flexible morphology. This allows for the creation of more emotionally colored and expressive linguistic constructions. ● Different Types of Thinking: The analytical approach of the English language reflects pragmatism and a desire for accuracy and specificity, while the synthetic approach of the Ukrainian language reflects a more holistic and figurative perception of the world. ● Values and Priorities: The emphasis on nouns in the English language reflects the focus on objectivity and the material world, while the greater focus on adjectives and verbs in the Ukrainian language reflects the value of dynamics, action, and expressiveness. ● Reflection of Cultural Characteristics: Differences in attitudes between compound word components reflect cultural characteristics, such as the English aim toward practicality and the Ukrainian tendency to imagery and emotionality. ● Different Ways of Categorization: Differences in the ways compound words are formed reflect different ways of categorizing and organizing information, characteristic of speakers of these languages. As we see, only one linguistic sphere (compound words) studied and contrasted between languages within one linguistic family reveals profound cognitive and cultural differences between speakers, allowing us to come closer to unveiling of real philosophical concepts. These distinctions reflect cognitive and cultural orientations. English prioritizes utility, precision, and objectivity, whereas Ukrainian compounds tend to foreground emotion, humor, spirituality, and holistic perception. These distinctions offer a window into the cognitive styles favored by each linguistic community and underscore how deeply language shapes the lived experience. 4.2 AI Language Output Divergence The multilingual AI output experiment revealed striking differences in how linguistic and cultural nuances are retained - or erased - depending on the language interface used. These discrepancies underscore how AI systems reflect the structure and limitations of their training data, often prioritizing dominant language frameworks, especially English, while failing to replicate the cognitive-emotional depth found in less represented linguistic cultures. English Output: Functional but Culturally Flat. When prompted in English, the AI-generated content emphasized structural characteristics of compounding: grammatical formation, productivity, stylistic usage, and semantic clarity. It introduced analytical categories such as loanword frequency , prevalence of compounding , and linguistic innovation tolerance . However, it failed to capture metaphorical, emotional, or culturally embedded features. Descriptions were technocratic, pragmatic. Emotional coloration, humor, or symbolic significance - central to the Ukrainian compounds - were omitted entirely. The English output mirrored the data-centric, utility-focused mindset of its linguistic base. Ukrainian and Polish Outputs: Surface-Level Cultural References. When prompted in Ukrainian or Polish, the AI produced outputs that shared a common structural pattern and vocabulary, suggesting mutual translation or mirroring. These responses acknowledged cultural associations with nature and traditional life (e.g., agriculture, folklore), but rarely moved beyond superficial description. Emotional nuance, irony, and metaphorical creativity were underrepresented. The outputs often felt encyclopedic rather than reflective of lived cultural experience. This likely results from the AI’s limited access to rich, diversified training data in these languages. Russian Output: Culturally Sensitive. The most culturally sensitive output came in response to the Russian prompt. The AI preserved categories such as emotional tone, metaphorical layering, humor, and evaluative framing. It recognized the pragmatic nature of English compounding and the emotive, holistic tendencies of Slavic compounds. The Russian-language response aligned closely with the findings of the author’s human-led contrastive analysis, indicating that Russian likely benefits from a more robust corpus of culturally relevant linguistic studies in the AI’s training data. Comparative Implications Across all outputs, the AI demonstrated a preference for structural and surface-level features when operating in or translating from English. Emotional, symbolic, and culture-specific meaning was most likely to appear in Slavic-language outputs and was almost entirely absent from English ones. Even when responding in other languages, the AI exhibited patterns consistent with English-centric cognitive models - suggesting English often serves as the implicit cognitive template for multilingual output. These results highlight a core limitation in the current generation of multilingual AI: the uneven representation of worldviews across languages. AI is not neutral - it is a product of its data environment. 5. Discussion The findings of this study highlight a crucial yet under examined tension at the intersection of artificial intelligence and linguistic diversity. While AI has become a powerful tool for global communication, content generation, and knowledge dissemination, it remains heavily shaped by the linguistic and cultural assumptions embedded in its training data. As a result, AI not only processes language -it privileges certain worldviews while marginalizing others. The contrastive analysis of English and Ukrainian compound formation revealed that these languages encode reality differently. English tends toward analytical precision and utilitarian framing, while Ukrainian favors holistic, emotionally charged, and metaphorically dense expressions. These differences are more than stylistic - they reflect divergent modes of perception, categorization, and meaning-making. However, when evaluated through AI outputs, these distinctions were inconsistently preserved. AI systems trained primarily on English data tended to reproduce English-centric patterns of thought, even when generating text in other languages. This phenomenon is a form of what is already termed linguistic imperialism via AI [ 6 ]. When AI systems default to English as their cognitive base, other languages, particularly those with fewer digital resources, are filtered through a narrow lens. Metaphors become flattened, emotions neutralized, and culturally resonant imagery replaced with generic formulations. The more synthetic, expressive, or context-dependent a language is, the more its unique features are likely to be misrepresented or ignored altogether. Moreover, this bias is not limited to translation. It appears in AI-generated content, learning applications, search results, and recommendation engines. As AI increasingly mediates our understanding of reality, its subtle cognitive filtering shapes not only how we communicate but what we imagine to be possible or real. This creates a feedback loop: the more AI is trained on dominant languages, the more those languages dominate the informational ecosystem, which accelerates the diminishing of linguistic and cultural diversity and at the same time it obstructs tools of self-reflection for dominant language culture. This raises pressing ethical concerns. Language is not merely a technical medium; it is the vessel of memory, identity, worldview, and community. The loss of linguistic nuance is a cognitive and cultural loss. For languages already endangered or digitally marginalized, the rise of AI presents both an opportunity and a threat: an opportunity to preserve and document, and a threat of homogenization if diversity is not actively supported. AI needs to be conceptualized and trained in a different way. It should be a multimind system , capable of accommodating and learning from a multiple of linguistic worldviews. This requires not only more diverse training data but fundamentally different design principles: AI must be taught to recognize that different languages encode different realities, and that cultural nuance is not noise, it is signal. This insight leads directly to the proposal outlined in the next section: the creation of an Atlas of Language Worldviews , a platform for capturing and amplifying linguistic diversity in the age of AI. 6. Proposal: Atlas of Language Worldviews 6.1. Concept and Applications In response to the challenges identified in the previous sections, an Atlas of Language Worldviews is suggested to be created - a multilingual, AI-assisted platform designed to preserve, compare, and promote the diverse cognitive and cultural frameworks encoded in language, as well as to train AI itself. The goal is not merely to document vocabulary or grammatical features but to map how different language communities perceive, categorize, and emotionally relate to the world. Creating an "Atlas of Worldviews" that engages AI and linguistic data to provide culturally-enhanced services and train AI seems to have a huge potential. The following scenarios would be possible using the Language Worldview platform: ● Language learning with cultural context (e.g., when and why certain phrases are used), ● Cross-cultural business : AI identifies cultural misunderstandings and suggests rapport-building strategies, ● Marketing & media : Flags culturally inappropriate content and suggests localized alternatives, ● Historical interpretation : Presents multiple cultural views on events. ● Communication aid : Suggests respectful phrasings and helps navigate cultural nuances, ● Social media analysis : Flags cultural friction and identifies trends through a worldview lens. The culturally-enhanced service the Atlas platform would result with an extended contextualized analysis of the prompt. For example, with the scenario of "culturological differences of compound word-making in English and Ukrainian", AI's Role can be to • to analyze the grammatical rules and patterns of compound word formation in English and Ukrainian (linguistic analysis) ; • to access the Atlas of Worldviews to identify relevant cultural values and beliefs associated with each language/culture ( cultural contexting) ; • to trace the historical development of compound word formation in each language, taking into consideration any cultural or social influences ( historical perspective) ; • to examine how the different ways of compounding words reflect different cognitive styles or ways of thinking(for example, more analytical vs. more holistic, more practical vs. more emotional) ( cognitive styles) ; • to analyze the metaphors embedded in compound words and how they reflect cultural values and beliefs ( metaphorical analysis) ; • to present the information in a clear and concise way, highlighting the cultural nuances and differences, to include references to relevant scientific studies and cultural texts ( holistic analysis) . The output wouldn't just be a linguistic comparison but a rich, contextualized analysis that reveals the underlying cultural assumptions and perspectives that shape language. 6.2. Roadmap of Language Woldview Atlas creation Let’s try to explore a practical roadmap for developing an Atlas of Worldview, outlining the key steps. 6.2.1. “Worldview” Definition The 1st step would be giving operational definition for the " worldview " notion for the Atlas. It is necessary to define what constitutes a "worldview" in a way that's measurable and can be mapped across cultures. This requires moving beyond general philosophical notions and identifying concrete, observable elements. Like, for example, values, morals, beliefs, social structures, cognitive styles, relationship to nature, etc. Each of the elements would be the answer to a set of categorizing questions. ● Values : What are the core values prioritized within a culture? (for example, individualism vs. collectivism, practicality vs. theorisation, civilizedness vs. pristineness, family values, environmental stewardship) How are these values expressed in language, traditions, and institutions?; ● Beliefs about Existence : What are the dominant beliefs about the nature of reality, the universe, and humanity's place in it? (for example, religious beliefs, spiritual practices, scientific understanding); ● Moral Frameworks : What is considered right and wrong? What are the ethical principles that guide behavior?; ● Social Structures : How is society organized (for example, hierarchies, kinship systems, gender roles, economic systems)?; ● Cognitive Styles : How do people perceive, process, and categorize information? (for example, holistic vs. analytical thinking, tolerance for ambiguity, communication styles); ● Relationship to Nature : How does a culture understand and interact with the natural world? (for example, stewardship, exploitation, reverence), etc. Once we have a working definition of "worldview," it would be necessary to identify observable markers . These markers will be the data points the AI would use. For example they may be: ● Metaphors, idioms, grammatical structures, common proverbs, the way time and space are conceptualized in language - linguistic features ; ● Art, music, literature, architecture, religious texts, cinema - cultural artifacts ; ● ◦Customs, traditions,rituals, etiquette, ethics, legal systems, political structures - social practices ; ● The stories a culture tells about itself, its origins, and its heroes - historical statements; ● The dominant intellectual frameworks used to understand the world - scientific and philosophical knowledge . (!) There should be some " Standard Worldview" anchor , like Greenwich prime meridian for the time zones. This anchor shouldn’t be a superior standard, but rather a reference point for comparison, like a baseline in a statistical analysis. Explicitly identifying it as a reference point allows users to see how other worldviews differ and relate to it. It acknowledges that Western perspectives have historically dominated global discourse and academic research. There is a challenge though that it could be misinterpreted as a normative standard, perpetuating Western biases. However it seems Greenwich prime meridian hasn’t had such interpretations for being a 0 longitude. Moreover, there is a huge worldview diversity within "Western" worldviews. It is not a monolithic entity. As we already noted, there are numerous variations across European countries, North America, etc. 6.2.2. Team The next step would be forming a multidisciplinary team : it would be necessary to bring together experts in linguistics, anthropology, AI, computer science, ethics, and cultural studies. 6.2.3. Research Plan Next, a detailed research plan would need to be developed: the specific research questions are to be outlined, as well as data sources, and methodologies that would be used. 6.2.4. Funding Funding needs to be secured: Funding opportunities from government agencies, foundations, and private investors are to be explored. 6.2.5. Data Sources One of the next steps presupposes specifying data and finding data sources : The AI needs to be trained on a large and diverse dataset of worldview-related information. There should be different types of data used: ● Linguistic Data: ○ Massive corpora of text and speech in multiple languages; ○ Grammatical structures, sentence patterns, idioms, metaphors, and slang; ○ Etymological information tracing the historical evolution of words and concepts. ● Cultural Data: ○ Historical events, religious beliefs, social norms, customs, artistic expressions, folklore, and traditions ( explicit cultural knowledge ); ○ Implicit cultural knowledge: values, assumptions, beliefs embedded in stories, proverbs, rituals, and social interactions ( explicit cultural knowledge ); ○ Gestures, body language, facial expressions, and their cultural connotations ( nonverbal communication data ); ○ Data on social hierarchical structures and power dynamics within different cultures. ● User interaction data (anonymized and aggregated): ○ Data on how users from different cultural backgrounds interact with the platform and its associated AI services. ○ Feedback on the accuracy and relevance of the platform's cultural insights. When we think of data sources : ● A natural starting point would be UNESCO's World Atlas of Languages , It could help to explore if language families could be used as proxies for worldview similarities; ● Ethnographic databases could be another resource, for example., databases like the Ethnographic Atlas, the Human Relations Area Files (HRAF), and similar resources, which contain a huge amount of information about cultural practices; ● Large collections of text and speech data ( linguistic corpora ) can be analyzed for patterns and insights into language use; ● Academic research data could be used to do research in anthropology, sociology, linguistics, psychology, philosophy, religious studies, and other relevant fields; ● Open data sets related to cultural indicators , social statistics, and global values surveys could be another resource; ● Crowdsourcing , an ethically-sourced and professionally validated community, can contribute to capturing nuances not found in formal datasets (for example, local proverbs, stories, customs). (!) It is necessary to be aware of potential biases in existing data. Many datasets over-represent Western perspectives or focus on easily accessible cultures. It makes sense to actively seek out data from underrepresented regions and perspectives and critically evaluate the sources and methodologies used to collect the data. Data Standardization is necessary to develop a system for coding and standardizing data from diverse sources so that it can be easily analyzed by the AI. 6.2.6 . Platform Architecture The next step would be Development of the AI Platform itself. ● It is necessary to choose an appropriate AI architecture . Possibilities include: NLP & ML - to analyze and learn cultural data ; deep learning - to infer complex worldview links; knowledge graphs - for relational worldview mapping; worldview modeling algorithms -for core belief simulation. ● It's crucial to make the AI's reasoning transparent. Users need to understand the reason the AI is providing a particular response. There should be methods developed for explaining the AI's decision-making process; ● Ethics: The AI should be designed to be fair, unbiased, and respectful of cultural differences. It should not utilize stereotypes or promote certain ideologies. The AI needs to be highly sensitive to cultural nuances and avoid making generalizations or assumptions. It should ensure that the AI is trained on data that accurately represents the diversity of worldviews. When using crowdsourced data, informed consent from contributors should be obtained and their privacy should be protected. The fact of technology misuse should be considered, and there should be safeguards implemented to prevent it. 6.2.7. Interface User-friendly interface ( both - AI training and end-user ones ) should be created, it should allow users to easily access and explore the Atlas of Worldviews. AI Training Interface should include; ● Tools for data scientists and cultural experts to annotate and validate the platform's knowledge; ● Visualization tools for exploring cultural differences and similarities; ● Debugging tools for identifying and correcting biases in the AI models. End-User Interface is to be featured with ● Culturally-adapted output: The AI returns responses tailored to the user’s perceived cultural background, using appropriate language ( culturally sensitive translation capability) , tone, and communication style. ● For more advanced users, the platform can provide a “cultural context ” panel alongside the AI's output. This panel might include: ○ Explanations of culturally specific terms or concepts used in the AI's response, ○ Information about potential cultural misunderstandings that could arise from the AI's output, ○ Alternative phrasings or communication strategies that might be more effective in certain cultural contexts, ○ Relevant historical or social background information. ● Ways are to be suggested to rephrase prompts to be more culturally sensitive or to elicit more relevant information from the AI ( prompt enhancement suggestions ). ● Visualization tools , like maps, charts, and other visual aids could help users understand complex data. ● Users could be allowed to explore the potential consequences of different cultural approaches to known values and beliefs ( interactive simulations ). 6.3. Language Worldview Atlas Training AI on Cultural Nuances 6.3.1. Data Structuring for AI Consumption The Atlas would collect and organize linguistic and cultural data into structured formats: ● Lexical data (e.g., culturally specific metaphors, idioms, compounds, neologisms) ● Cognitive patterns (e.g., how time, space, emotion are perceived in a culture) ● Pragmatic conventions (e.g., politeness, indirectness, use of diminutives) ● Moral and value frameworks (e.g., collectivism vs. individualism) These would be annotated and tagged (e.g., "Ukrainian_holophrastic_compounds," "English_efficiency_bias," "Mandarin_vertical_time"). 6.3.2. Integrating with Language Models The structured data can then be used to fine-tune existing LLMs or to train culturally aware models from scratch: ● Fine-tuning: Inject worldview-tagged texts into models like GPT, LLaMA, or Mistral to increase sensitivity to specific cultural patterns. ● Control tokens: Embed “cultural worldview prompts” to guide generation—e.g., <>, <>. ● Retrieval-Augmented Generation (RAG): Use the Atlas as a semantic knowledge base to supplement AI responses dynamically during generation. 6.3.3. Cross-Cultural Evaluation Tasks The Atlas would support the creation of benchmark tasks that test: ● Can the AI generate culturally nuanced explanations? ● Can it switch metaphors appropriately across languages? ● Can it recognize worldview conflicts in multilingual input? This allows evaluation and feedback loops , improving the model's ability to emulate culturally embedded reasoning. 6.3.4. Bias Detection & Debiasing AI systems often reflect Western-centric biases in: ● Moral framing ● Emotional tone ● Examples and assumptions The Atlas can help audit AI outputs and reveal worldview imbalances by: ● Comparing outputs across worldview categories ● Flagging culturally inappropriate translations or generalizations 6.3.5. Training AI for Intercultural Mediation The Atlas could train AI assistants to: ● Detect worldview mismatches in cross-cultural communication ● Suggest culturally respectful phrasings ● Offer meta-comments like: “Note: this phrase may carry a different tone in Ukrainian than in English.” No doubt, if started small, the project may be started sooner. If we begin with a pilot project focusing on a limited number of cultures and worldview elements, and continuously evaluate and refine the Atlas of Worldviews based on user feedback and research findings, we can create a valuable resource that promotes intercultural understanding and transforms the way we interact with the world. 7. Conclusion As artificial intelligence continues to enter communication, education, governance, and cultural production, it is essential to recognize that AI reflects the language models it is trained on - and thus the worldviews those languages encode. If AI is trained primarily on English and other dominant global languages, it will inevitably replicate the cultural and cognitive assumptions embedded within them, often at the expense of linguistic and cultural diversity. This paper has illustrated how even within the same language family, profound cognitive and emotional differences exist in how speakers of English and Ukrainian percept and categorize knowledge. It has further shown that current AI systems often fail to replicate these subtleties, especially when dealing with less-resourced languages. What is lost in this process is not just linguistic detail, but a deeper capacity for comparative self-understanding, intercultural empathy, and meaningful communication. Language is not just a code for information - it is a lens through which reality is perceived, remembered, and shaped. When that lens is narrowed by technology, so too is our collective imagination. If we allow AI to standardize thought, we risk silencing the plurality of human experience. The proposed Atlas of Language Worldviews offers a concrete and scalable response to this challenge. It recognizes that each language is a repository of culturally specific cognitive patterns and that AI systems must be trained to honor this diversity rather than override it. Such a platform would not only support the preservation of endangered languages but also contribute to the creation of AI that is more ethical, responsive, and genuinely global. To conclude, it is necessary to outline the main opportunities of the Woldview Atlas. It can become an enhanced tool for the following: • understanding and appreciating different worldviews (promoting intercultural understanding); • facilitating more effective communication across cultures (improving communication); • providing students with a deeper understanding of cultural diversity (enhancing education); • moving beyond simple translation to culturally-aware nuanced AI assistance (revolutionizing AI services); • facilitating more effective collaboration on global challenges (supporting global collaboration); • better negotiations, stronger relationships with international partners, and more successful marketing campaigns (more effective global business); • by promoting understanding and mitigating bias, the platform could contribute to creating more inclusive and equitable societies; • innovation in fields relating to languages, cultural sciences, philosophy, sociology, communication and global studies. Ultimately, the project is about expanding the scope of what we, as humans, can know about ourselves, about others, and about the world we share. In a time when digital systems are reshaping reality, ensuring that cultural multiplicity survives and thrives is not an academic luxury, it is a human necessity. References Бугров ВА (1996) Проблема мови у творчості пізнього Л.Вітгенштейна. Язык и культура: 4-я Междунар. конф.: Материалы., vol. 1, pp. 268–277 Винокур ГО (1959) Заметки по русскому словообразованию. Избранные работы по русскому языку ed., Москва Омельченко ЛФ (1981) Продуктивные типы сложных слов в современном английском языке (на материале прилагательных и глаголов. Киев, Вища школа. Головное изд-во Серебренников БА, Уфимцева AA (eds) (1977) Языковая номинация: Общие вопросы. Москва, Наука Allard-Kropp (2020) Manon. Languages and Worldview. St. Louis, University of Missouri–St. Louis, Open Educational Resources (OERs). https://irl.umsl.edu/oer/17 . Accessed 25 march 2025 Arun C (2024) Transnational AI and Corporate Imperialism. Carnegie Endowment for International Peace, https://carnegieendowment.org/, https://carnegieendowment.org/research/2024/10/transnational-ai-and-corporate-imperialism?lang=en . Accessed 25 March 2025 Boroditsky (2019) Lera, director. How language shapes the way we think. , https://www.youtube.com/watch?v=RKK7wGAYP6k. Accessed 25 March 2025 Boroditsky (2021) Lera, director. How Language Shapes Thought. , https://www.youtube.com/watch?v=I64RtGofPW8. Accessed 25 March 2025 Boroditsky L et al (2011) Do English and Mandarin speakers think about time differently? Cognition, vol. 118, no. 1, pp. 123–129. https://www.sciencedirect.com/ , https://doi.org/10.1016/j.cognition.2010.09.010 . Accessed 25 March 2025 Cook VJ, Newson M (2007) Chomsky's universal grammar: an introduction. Wiley Davis G (2024) GPT-4o and Emotion Recognition Dramatically Changes AI Human Interaction. Intelligentliving. https://www.intelligentliving.co/gpt-4o-emotion-recognition-ai-human/ . https://www.intelligentliving.co/, https://www.intelligentliving.co/gpt-4o-emotion-recognition-ai-human/. Accessed 25 March 2025 Jonathan Winawer J et al (2007) Russian blues reveal effects of language on color discrimination. https://www.pnas.org/ , vol. 104, no. 19, PNAS. https://www.pnas.org/doi/full/10.1073/pnas.0701644104. Accessed 25 March 2025 Kelly L (2019) 50 Words For Snow. https://readable.com/blog/, 8 Feruary https://readable.com/blog/do-inuits-really-have-50-words-for-snow/ . Accessed 25 March 2025 Kornai András (2013) Digital Language Death. PLOS, https://plos.org/#journals. https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0077056 . Accessed 25 March 2025 Mair V (2023) Where did the PIEs come from; when was that? Language Log, no. Historical linguistics, Language and archeology, Language and genetics. https://languagelog.ldc.upenn.edu/ , https://languagelog.ldc.upenn.edu/nll/?p=59908. Accessed 25 March 2025 Malotki E (1983) Hopi time: a linguistic analysis of the temporal concepts in the Hopi language. Mouton, https://ebin.pub. https://ebin.pub/hopi-time-a-linguistic-analysis-of-the-temporal-concepts-in-the-hopi-language-9783110822816-9027933499-9789027933492.html . Accessed 25 March 2025 Mann H (2024) AI Homogenization Is Shaping The World. https://www.forbes.com /, 5 March https://www.forbes.com/sites/hamiltonmann/2024/03/05/the-ai-homogenization-is-shaping-the-world/. Accessed 25 March 2025 Marsden H (2024) https://theweek.com/tech/how-technology-helps-and-harms-endangered-languages . The Week. www.theweek.com, https://theweek.com/tech/how-technology-helps-and-harms-endangered-languages. Accessed 25 March 2025 Melero M. The Future of Language (and Cultural) Diversity in the Age of AI. CLARIN Annual Conference Barcelona15 October 2024. CLARIN, Barcelona Barcelona Supercomputing Center. https://www.clarin.eu/sites/default/files/01_Melero_CLARIN2024_Keynote_The-Future-of-Language-Diversity-in-the-Age-of-AI.pdf . Accessed 25 March 2025 NorthernLife N (2024) https://northernlifemagazine.co.uk/neologisms-2024 /, 26 November 2024. Núñez RE, Sweetser E (2006) With the Future Behind Them: Convergent Evidence FromAymara Language and Gesture in the CrosslinguisticComparison of Spatial Construals of Time. Cognitive Science, vol. 30, no. 3, pp. 401–450. https://onlinelibrary.wiley.com/doi/10.1207/s15516709cog0000_62 . Accessed 25 March 2025 Spier L et al (eds) (1960) Language, Culture, and Personality: Essays in Memory of Edward Sapir. Greenwood Press. https://books.google.com/books/about/_Culture_and_Personality.html?id=UbJxAAAAMAAJ . Accessed 23 March 2025 UNESCO Headquarters in Paris (2025) Language technologies for all - Advancing Humanism through Language Technologies. lt4all2025. https://lt4all2025.sciencesconf.org/ . Accessed 25 March 2025 Whorf BL (1956) An American Indian model of the universe. Language, thought, and reality: selected writings of Benjamin Lee Whorf, no. Language and languages, pp. 57–64 Whorf BL (1941) The Relation of Habitual Thought and Behavior to Language. https://is.muni.cz /, https://is.muni.cz/el/1423/podzim2006/SAN205/um/duranti_la_whorf.pdf. Accessed 25 march 2025 Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6719644","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Systematic Review","associatedPublications":[],"authors":[{"id":460174004,"identity":"5a3f6acd-21e1-487c-af6d-ff01371ebace","order_by":0,"name":"Tetiana Ilman","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAzUlEQVRIiWNgGAWjYJCCA0DIwA9iJRQQoZwHpkWyAaTFgEgtIE0MBgdADGK02LMff3jgw5nD8sbnVyd+eGDAIM8vdoCALTw5Bgdn3DhsuO3G280SQIcZzpydQMhhOQyHeT4cZtx24+wGkJYEg9uEtPA/fwDSYr95xtnNP4jTAjT5MM+Nw4kb+Hu3EWnLjTdAv5xJT55xg3ebRYKBBGG/sPenP/7w4Zi1bX//2c03f1TYyPNLE9CCABJglRLEKgcB/gOkqB4Fo2AUjIKRBAA9vUvxxsDYbQAAAABJRU5ErkJggg==","orcid":"","institution":"","correspondingAuthor":true,"prefix":"","firstName":"Tetiana","middleName":"","lastName":"Ilman","suffix":""}],"badges":[],"createdAt":"2025-05-21 22:23:17","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-6719644/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6719644/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":83344419,"identity":"fcc8c45a-8707-45fc-be52-b22fa58870fb","added_by":"auto","created_at":"2025-05-23 11:46:49","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2165463,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6719644/v1/1a8038de-85e8-45ac-942c-45a8ce28af17.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eAI's Blind Spot: Language Worldviews to the Rescue\u003c/p\u003e","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eMy fascination with the link between language and culture began at the age of six, with a Russian book The Tales of the Peoples of the World. I was struck by how vastly the stories differed from familiar Russian and Ukrainian ones - especially those from Australia, Oceania, eastern tribes. Names, humor, life goals, and values - all felt alien yet captivating. For me, that foreignness manifested in the strange names they gave to people, objects and phenomena, in the way they spoke. Even then, I could sense the connection between the language and their worldview.\u003c/p\u003e\n\u003cp\u003eThis early intuition evolved into an academic pursuit. \u0026nbsp;I began to explore an ancient yet unresolved question: do cultural/environmental differences cause languages and worldviews to differ, or vice versa: maybe the language is some type of encoded frame that shapes human cognition and worldview?\u0026nbsp; \u0026nbsp;This question has fascinated thinkers for centuries, from Plato, St. Augustine, Humboldt, Boas, Sapir, Whorf, Wittgenstein, Vygotsky to \u0026nbsp;modern cognitive linguists (Lakoff, Boroditski, Turbayne), specialists in language ideology, language behaviour, etc. Having reconsidered early ideas of linguistic relativity, contemporary studies across psycholinguistics, cognitive science, and anthropology continue to demonstrate that language reveals deep cognitive and cultural differences - \u0026nbsp;what we may call language-specific worldviews. It \u0026nbsp;influences how we perceive time, space, emotion, and social relationships and build our behavior and knowledge categorization. In my work I provide impressive worldview diversity examples from contemporary \u003cstrong\u003elinguistic relativity\u003c/strong\u003e studies as well as conduct my own research in how language structures (on the example of English and Ukrainian suffix derivatives and compounds) reveal embedded differences in cultural cognition. Derivation and compound word formation was not a randomly chosen sphere of contrastive worldview analysis. \u0026nbsp;In human speech, naming activity is the most telling aspect from the point of view of the correlation of language units with the extralinguistic objective world. \u0026nbsp; Naming in language is always meaningful\u003cstrong\u003e.\u003c/strong\u003e As languages of the Indo-European family, English and Ukrainian are both productive in forming compound words. Therefore, the c\u003cstrong\u003eontrastive analysis\u0026nbsp;\u003c/strong\u003eof English and Ukrainian worldviews forms\u003cstrong\u003e\u0026nbsp;the first line of research\u003c/strong\u003e.\u003c/p\u003e\n\u003cp\u003eThe idea of the second line of research was motivated by the \u0026nbsp; paradoxical situation we are facing. On one hand, \u003cstrong\u003eartificial intelligence\u003c/strong\u003e - especially large language models (LLMs) - has revolutionized access to knowledge, accelerating learning and linguistic processing at an unprecedented scale. On the other hand, these technologies pose a serious threat to linguistic and cultural diversity, which causes\u003cstrong\u003e\u0026nbsp;AI homogenization\u003c/strong\u003e. AI systems, often trained primarily on English data, are not culturally neutral. They encode and reproduce dominant cultural perspectives, flattening nuanced worldviews into standardized outputs. Many languages have not yet crossed the digital divide, and may never do so unless conscious steps are taken.\u003c/p\u003e\n\u003cp\u003eWe can state that we live in an era where we have not yet fully understood the worldviews encoded in languages, and at the same time we are already at risk of losing those nuances, which means we are losing a crucial tool of self-reflection.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eIn our world cognition advances through comparison.\u0026nbsp;\u003c/strong\u003eWithout contrasting categories from different cultures and languages, human understanding becomes narrower and less reflective. Comparing the culture of the language group to other cultures is an efficient tool for \u003cstrong\u003eself-reflection.\u0026nbsp;\u003c/strong\u003eWhen cultures examine themselves in the mirror of another language, they not only refine their self-awareness, but also reduce bias, improve communication, strengthen identity and better preserve valuable traditions. Without such reflection, cultural assumptions remain unchallenged and unexamined. The loss of a culture or a language is seen as not merely the disappearance of another worldview system but rather as the loss of a unique lens of world perception.\u003c/p\u003e\n\u003cp\u003eTo illustrate this problem, a contrastive analysis of English and Ukrainian worldviews is held along with the \u003cstrong\u003esecond line of research\u003c/strong\u003e: an \u003cstrong\u003eexperimental investigation\u003c/strong\u003e into how multilingual AI tools reflect (or fail to reflect) these linguistic worldviews in generated content.\u003c/p\u003e\n\u003cp\u003eThe findings reveal a critical \u003cstrong\u003eblind spot in AI development - \u0026nbsp;its insensitivity to the cultural and cognitive dimensions encoded in language\u003c/strong\u003e.The study explores how AI\u0026rsquo;s reliance on English-centered data and its preference for efficiency over nuance flattens cultural distinctions.\u003c/p\u003e\n\u003cp\u003eAs a solution, I suggest the development of a \u003cstrong\u003eLanguage Worldview Atlas\u0026nbsp;\u003c/strong\u003e- a multilingual platform for documenting, comparing, preserving cultural perspectives across languages, as well as for AI training. This project aims not only to address worldview erasure but to create a foundation for ethical and culturally aware AI systems.\u003c/p\u003e"},{"header":"2. Theoretical Background","content":"\u003ch2\u003eLinguistic relativity\u003c/h2\u003e\n\u003cp\u003eThe idea that language shapes thought has a long intellectual history, touching philosophy, anthropology, and modern cognitive science. Early thinkers such as Plato, St. Augustine, and Roger Bacon considered language to be central to understanding reality. In the 18th and 19th centuries, Wilhelm von Humboldt and Johann Gottfried Herder proposed that language expressed the \u0026quot;spirit\u0026quot; of a people - the insight would later influence the theory of linguistic relativity.\u003c/p\u003e\n\u003cp\u003eIn the early 20th century, this line of thought gained momentum through the work of Franz Boas and his students, notably Edward Sapir and Benjamin Lee Whorf. Whorf studied the language of Hopi (one of the Native American tribes) and suggested it had a different concept of time than \u0026ldquo;Standard Average European\u0026rdquo; \u0026nbsp;[\u003cem\u003e25\u003c/em\u003e] \u0026nbsp; speakers had. He gave arguments that Hopi had \u0026quot;no words, grammatical forms, construction or expressions that refer directly to what we call \u0026apos;time\u0026apos; and concluded that the Hopi had \u0026quot;no general notion or intuition of time as a smooth flowing continuum in which everything in the universe proceeds at equal rate, out of a future, through the present, into a past.\u0026quot; \u0026nbsp; [24] These particular ideas for time category in Hopi were later \u0026nbsp; refuted \u0026nbsp;[16]. However, the theory of linguistic relativity, often referred to as the Sapir-Whorf hypothesis, has modified and become a reliable reference for many interdisciplinary studies, like psycholinguistics, cognitive linguistics, behavioral linguistics.\u003c/p\u003e\n\u003cp\u003eContrastive psycholinguistic studies have shown that speakers of different languages conceptualize time, space, and color in ways aligned with the structure of their native tongue, for example, \u0026nbsp;that timelines come in all shapes and directions. For the Aymara people of the Andes, time flows front to back. The past, which was known and hence seen, lies in front. The unknown and unseen future is behind [21]. Mandarin speakers (China) sometimes represent time along a vertical axis, the past is above and the future is below [9].\u003c/p\u003e\n\u003cp\u003eAccording to new studies in language ideology [5], in the field of linguistic anthropology cultures can be divided into two groups based on the way that members of that culture generally perceive time: monochronic cultures and polychronic cultures. Cultures, called monochronic, tend to believe that time is linear. \u0026nbsp;They like to do one thing at a time, stick to schedules, and value being on time (as in the U.S., Germany, and Great Britain). Other cultures, called polychronic, tend to see time as fluid and malleable, they are more flexible with time, don\u0026apos;t mind interruptions, and prioritize relationships over strict schedules (as in Mexico or Egypt). Monochronic cultures value efficiency, while polychronic cultures are more adaptable and focus on people. In monochronic cultures these behaviors are thought to be inefficient and improper. People of polychronic cultures are more susceptible to distractions and open to interruption but are better at focusing on many tasks at once.\u003c/p\u003e\n\u003cp\u003eWittgenstein\u0026rsquo;s famous dictum -\u0026ldquo;The limits of my language mean the limits of my world\u0026rdquo;[1] - means that the scope of our understanding and experience of the world is fundamentally shaped and constrained by the language we use to describe and think about it, it captures the philosophical essence of linguistic relativity. If a concept cannot be named in one\u0026rsquo;s language, it becomes harder to notice, describe, or even think about. This is not merely a limitation of vocabulary but of conceptual worldview.\u003c/p\u003e\n\u003cp\u003eEmpirical evidence supports this notion. A well-known 2007 study on color discrimination revealed that Russian speakers \u0026nbsp;(whose language \u003cstrong\u003emake obligatory distinction\u003c/strong\u003e between light blue -\u0026quot;goluboy\u0026quot;- and dark blue -siniy\u0026quot;) were not only \u0026nbsp;faster at distinguishing between these shades than English speakers (whose language uses the single word \u0026ldquo;blue\u0026rdquo; for both), but \u003cstrong\u003erather that Russian speakers \u003cu\u003ecouldn\u0026rsquo;t avoid\u003c/u\u003e distinguishing them\u0026nbsp;\u003c/strong\u003e[12]\u003cstrong\u003e.\u0026nbsp;\u003c/strong\u003eThis communicative requirement appears to cause Russian speakers to habitually make use of this distinction even when performing a perceptual task that does not require language.The study suggested that linguistic representations normally interfere in objective perceptual decisions.\u003c/p\u003e\n\u003cp\u003eOther striking examples come from the language of the Kuuk Thaayorre people in Australia. They use cardinal directions (north, south, east, west) instead of \u0026nbsp;\u0026ldquo;left\u0026rdquo; and \u0026ldquo;right\u0026rdquo;. It shows that their spatial perception is rather Earth-oriented, unlike ours - self-oriented. \u0026nbsp;A speaker might say, \u0026ldquo;There\u0026rsquo;s an ant on your southwest leg\u0026rdquo;.This linguistic habit cultivates exceptional spatial orientation skills. \u0026nbsp;\u0026quot;Hello\u0026quot; in Kuuk would literally mean, \u0026quot;Which way are you going?\u0026quot;. The answer could be, \u0026quot;North-northeast in the far distance. How about you?\u0026quot;\u003c/p\u003e\n\u003cp\u003eThis constant need to stay oriented allows speakers of this language to maintain a strong sense of direction, even without special gadgets, like compasses, demonstrating that humans are capable of much better orientation than previously thought if trained by their language and culture.\u003c/p\u003e\n\u003cp\u003eThere is a distinctive peculiarity in how the speakers of \u0026nbsp;Kuuk Thaayorre Aboriginal group organize time based on cardinal directions. When asked to arrange pictures of a person aging, they do not order the sequence from left to right or right to left (as English or Hebrew speakers might). Instead, they orient the timeline based on the direction they face, suggesting that their sense of time is landscape-centered rather than self-centered. Indeed, why would we organize time self-centered if we remember that the sun rises on the east and goes down on the west? Must be so selfish from Kuuk\u0026rsquo;s point of view.\u003c/p\u003e\n\u003cp\u003eThe Holy Roman Emperor, Charlemagne, said, \u0026ldquo;To have a second language is to have a second soul\u0026rdquo;, this is a strong statement supporting the idea that \u0026nbsp;language crafts reality. \u0026nbsp;Similarly, \u0026ldquo;as many languages you speak, as many times you are human/ the more languages you speak, the more times you are human\u0026rdquo;. The sayings are attributed to different people - Chekhov, Goethe, Masaryk. All these sayings mean that learning a new language is about entering a new culture, understanding different ways of thinking, and seeing the world through another lens. \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003ch2\u003e2.1.From Chomsky\u0026rsquo;s Universal Grammar to AI \u0026ldquo;imperialism\u0026rdquo;\u003c/h2\u003e\n\u003cp\u003e\u0026nbsp;The rise of cognitive science in the latter half of the 20th century has shifted the focus of studies from semantics to syntax, to generative approach. Noam Chomsky pioneered the hypothesis \u0026nbsp;that syntax was the level reflecting a structure of the human mind. Chomsky\u0026apos;s Universal Grammar (or General Grammar) theory developed primarily by Chomsky [10] \u0026nbsp; suggested that every person has an innate mental structure of production and understanding of language, a set of abstract principles and parameters that underlie all human languages, or \u0026quot;universal grammar\u0026quot;.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe goal was to make a complete model of the inner language, and the model could then be used to describe all human language and to predict if any utterance would sound correct to a native speaker of the language. The Universal Grammar theory has provided the instrumental for AI creation, particularly in the areas of natural language processing (NLP) and understanding, as well as machine learning algorithms, which are a critical part of many AI systems. These algorithms are designed to recognize patterns and make predictions based on data. It provided a foundation for the development of AI systems that can process, understand natural language and respond to human language in a more sophisticated, \u0026nbsp;natural, intuitive and human-like way.\u003c/p\u003e\n\u003cp\u003eAI appeared in 2023 and its development has been explosive. \u0026nbsp;It has been changing the world since then. \u0026nbsp;It has turned into a revolutionizing instrument used at every level of human life. AI-centered scientific investments prevail considerably at this time, and this tendency will most likely only accelerate. World AI leading countries take part in the race for AGI (\u003cstrong\u003eArtificial general intelligence\u003c/strong\u003e, which is a type of AI that matches or surpasses human cognitive capabilities across a wide range of cognitive tasks).\u003c/p\u003e\n\u003cp\u003eScience these days is trying to figure out how \u0026ldquo;homogenezative\u0026rdquo; AI can be for culture and creativity, and language diversity [17]-[19]. AI \u0026nbsp; has the potential to diminish culturological differences between languages and \u0026nbsp;creativity: The main reasons for this to happen can be identified as:\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eCoding Languages and \u0026nbsp;Terminology\u003c/strong\u003e:\u003c/p\u003e\n\u003cp\u003e○ \u003cstrong\u003e\u003cem\u003eEnglish is the Dominant Language\u003c/em\u003e\u003c/strong\u003e: English has been the dominant language on the internet since its beginning, largely due to its widespread use in business, education, and technology. English has been Lingua Franca (dominating) \u0026nbsp;geographically and across domains. Almost everything we know now about the human mind and human brain is based on studies of usually English-speaking \u0026ldquo;undergraduates at universities\u0026rdquo; [7].\u003c/p\u003e\n\u003cp\u003eMuch of the codebase, documentation, and technical literature in the AI and technology world is in English. Even as AI is used to translate and localize these resources, it still subtly reinforces the dominance of English in \u0026nbsp; these domains. Users of AI, developers or end-users, are frequently exposed to English terminology, subtly shaping the way they think about technology and its potential. This can impact how other languages are used to describe similar concepts, potentially borrowing English terms rather than developing native ones.\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eTranslation/ Interpretation\u003c/strong\u003e:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e○ \u003cstrong\u003e\u003cem\u003eDominance of Certain Languages\u003c/em\u003e\u003c/strong\u003e: Current translation AI models are often trained more extensively on English, major languages, like Mandarin, and Spanish. This can lead to translations into these languages being of higher quality and easier to understand. Over time, individuals may be subtly predisposed to communicate primarily in these dominant languages to ensure their message isn\u0026apos;t lost in translation, further marginalizing less common languages and their associated cultures.\u003c/p\u003e\n\u003cp\u003e○ \u003cstrong\u003e\u003cem\u003eHomogenization of Meaning\u003c/em\u003e\u003c/strong\u003e: AI-powered translation tools, even though it has been constantly improving, can sometimes flatten nuances, idioms, and cultural references that are deeply embedded in a language. They might prioritize conveying the core meaning but miss the subtle undertones, emotional weight, or historical context that make a phrase or expression unique in its original language. Over time, reliance on these tools could lead to a standardization of language, where complex culturally-specific terms are replaced with more universally understandable alternatives.\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eContent Creation and Consumption\u003c/strong\u003e:\u003c/p\u003e\n\u003cp\u003e○ \u003cstrong\u003e\u003cem\u003eAI-Generated Content in Other Languages\u003c/em\u003e\u003c/strong\u003e: AI can generate content (text, images, video, etc.) in various languages. However when the AI is trained primarily on a single cultural dataset (likely a Western dataset), it might implement biases and cultural assumptions into the content it produces, even when generating in a language associated with a different culture. \u0026nbsp; This is a form of cultural \u0026ldquo;imperialism\u0026rdquo; via AI [6]. This could lead to a convergence toward a more globalized, homogenized cultural aesthetic and value system.\u003c/p\u003e\n\u003cp\u003e○ \u003cstrong\u003e\u003cem\u003ePersonalized Recommendation Systems\u003c/em\u003e\u003c/strong\u003e: AI systems already \u0026nbsp;that recommend books, movies, and music from around the world. While it can be a great tool for cultural exchange, if these systems are designed to enhance engagement, they push users toward content that aligns with their pre-existing preferences (often shaped by their local culture). This could reduce exposure to diverse cultural viewpoints and expressions, reinforcing existing echo chambers and inhibiting true cross-cultural understanding. The Forbes article \u0026quot;AI Homogenization Is Shaping The World\u0026quot; by Hamilton Mann (Forbes, 2024) [17] argues that the increasing reliance on a small number of large AI models, particularly from major tech companies, is already leading to a homogenization of thought and creativity.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eLanguage Learning\u003c/strong\u003e:\u003c/p\u003e\n\u003cp\u003e○ \u003cstrong\u003e\u003cem\u003eStandardized Language Instruction\u003c/em\u003e\u003c/strong\u003e: AI-powered language learning tools can be incredibly effective, but they can also promote a standardized version of a language. They may focus on grammar and vocabulary that are considered \u0026quot;correct\u0026quot; or \u0026quot;common\u0026quot; potentially neglecting regional dialects, slang, and other forms of linguistic diversity that are essential facets of a culture.\u003c/p\u003e\n\u003cp\u003e○ \u003cstrong\u003e\u003cem\u003eLoss of Cultural Immersion\u003c/em\u003e\u003c/strong\u003e: While AI-tutoring can enhance learning, it cannot fully replicate the experience of cultural immersion that is crucial for understanding the subtle nuances and context behind a language\u0026apos;s usage. Learning a language is more than memorizing words; it\u0026apos;s about understanding the cultural context from where \u0026nbsp;it springs.\u003c/p\u003e\n\u003cp\u003eThe sad news is that linguistic studies show that, of the approximately 7,000 languages spoken today, about 2,500 are generally considered endangered. \u0026ldquo;Less than 5% of all languages can still ascend to the digital realm\u0026rdquo; [14]. \u0026nbsp;Digital diglossia puts low-technological languages at risk of extinction.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;Many languages haven\u0026rsquo;t yet crossed the digital divide for many reasons. Some of them have little or no Internet access; some of them do not have a well-developed research community, some don\u0026rsquo;t get support from public administrations, etc.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eYoung speakers of minor languages are growing up in conditions with the Internet and a global world with languages like English, Spanish, Chinese and other major languages being pervasive for virtual communication, digital and social media. Didactic materials in major languages are effortlessly available over and above those of native minor languages.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe risks have been paid attention to. Since 2019 Unesco Headquarters in Paris have been holding conferences devoted to Language technologies (Language Technologies for All - LT4All 2019). The 1st one, \u0026nbsp;themed \u0026ldquo;Multilingualism for Building Knowledge Societies\u0026rdquo;, highlighted the critical role of language and cutting-edge technology, including artificial intelligence, in shaping cross-cultural communication. \u0026nbsp;The conference resulted in certain initiatives by research institutions and technological companies toward developing language technologies for a wider range of languages. \u0026ldquo;Despite significant progress, however, many communities are still being left behind\u0026rdquo; [23].\u003c/p\u003e\n\u003cp\u003eThe 2nd International Conference \u0026nbsp;(LT4All 2025) themed \u0026ldquo;Advancing Humanism through Language Technologies\u0026rdquo; and aimed at furthering the agenda of language technologies with a focus on community empowerment [19]. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAt the LT4All Conference in March 2025, it has been reported that even though:\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; Large multilingual datasets are being created (like The OSCAR project - Open Super-large Crawled Aggregated coRpus)\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;LLMs keep getting larger and (some of them) more diverse\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Meta developed the first multilingual translation models that don\u0026rsquo;t rely on English as pivot (2020-2022)\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Generative AI can create content in different languages from the start, dispensing with the traditional translation and localization process.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSTILL, HOWEVER\u003c/strong\u003e the future of language (and cultural) diversity in the age of AI is uncertain and the challenges should be addressed with proper awareness and action on different levels: \u0026nbsp;the language \u0026nbsp;and \u0026nbsp; the research communities, governments and public institutions and all of us.\u003c/p\u003e"},{"header":"3. Methodology","content":"\u003ch2\u003e3.1 Linguistic Contrastive Analysis\u003c/h2\u003e\n\u003cp\u003eThis study draws on an extensive contrastive analysis of word formation in English and Ukrainian. The research approach is grounded in the assumption that naming practices - especially compounding - reflect how speakers of different languages conceptualize the world.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eNaming is always mediated by thought. \u0026nbsp;A compound word, as a unique unit of nomination, primarily indicates that it is enough for a speaker to name two related objects for the listener to independently guess the type of connection the speaker had in mind.\u003c/p\u003e\n\u003cp\u003eUnderstanding the patterns of nomination leads to a deeper understanding of the role of the human factor in language. Any complex word can be considered a frozen context for its elements, therefore, analyzing those compound words in English and Ukrainian promise to reflect national specificities in perception and understanding of the world.\u003c/p\u003e\n\u003cp\u003eA new derivative, being a creation of word-formation act, like any new word, represents the product of the generalizing mental activity of a person, a product of understanding the features of the object being named, a result of singling out one feature from many, and thus a result of abstraction. After all, it is undoubtedly that \u0026nbsp;the derivation process and compounding (these \u0026nbsp; processes that can be directly observed in living speech) do reflect, as a mirror, numerous characteristic features of the act of nomination. Thus, it\u0026apos;s generally believed that \u003cstrong\u003emost, if not all, natural languages have some form of compounding.\u003c/strong\u003e Compounding is observed across different language families and linguistic typologies.\u003c/p\u003e\n\u003cp\u003eParticular attention in the work was devoted to observing the semantic connections between objects of reality in the process of word-formative nomination in English and Ukrainian and, with its help, \u0026nbsp;there was an attempt to compare two linguistic \u0026quot;worldviews\u0026quot; (understandably, fragmentary and illustrative) - English and Ukrainian. The \u003cstrong\u003econtrastive method\u0026nbsp;\u003c/strong\u003ein the research was used in combination with the \u003cstrong\u003etechnique of decomposition or reconstruction of the word-formation act\u003c/strong\u003e. \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBoth languages studied, English and Ukrainian, form compounds using similar word classes (nouns, adjectives, verbs, adverbs). \u0026nbsp; As languages of the Indo-European family, they are both productive in forming compound words.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe analysis was based on data drawn from multiple sources, including general and specialized dictionaries, literary texts, journalistic materials, and examples from everyday speech. Emphasis was placed on compounds that are lexicographically fixed, emotionally marked, or culturally embedded.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe analysis focused on three semantic fields:\u003c/p\u003e\n\u003col start=\"1\" type=\"1\"\u003e\n \u003cli\u003e\u003cstrong\u003eNature and Environment Vocabulary\u003cbr\u003e\u003c/strong\u003e Compound names related to flora, fauna, geography, and the domestic environment. These terms often reveal the associative patterns speakers use to connect physical traits with metaphorical or\u0026nbsp; functional significance.\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eAnthropocentric Vocabulary\u003cbr\u003e\u003c/strong\u003e Compounds that describe human attributes, behaviors, or roles. These often rely on metaphor, tactile imagery, and evaluative connotations. Attention was given to how attributes like strength, intelligence, \u0026nbsp;or emotionality are encoded differently across languages.\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eCulturological Vocabulary (Culturemes)\u003cbr\u003e\u003c/strong\u003e Words that reflect socially specific concepts, practices, or values. These include idiomatic formations, culturally loaded metaphors, and traditional expressions that encapsulate the worldview of a \u0026nbsp;linguistic community.\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eUkrainian, a synthetic language, relies more heavily on suffixation for emotional and diminutive expression, while English, an analytic language, leans on compounding for pragmatic categorization. The comparative study examined not only the structure of compounds but also the emotional, metaphorical, and evaluative layers embedded in them. Special attention was paid to diminutive-hypocoristic forms, imperative holophrases (especially in surnames and nicknames), and culturally resonant occasionalisms (creative neologisms). The goal was to reveal not just linguistic differences, but underlying cultural tendencies.\u003c/p\u003e\n\u003ch2\u003e3.2 Multilingual AI Output Evaluation\u003c/h2\u003e\n\u003cp\u003eTo complement the linguistic analysis, a second research line focused on testing how AI systems reproduce culturally embedded distinctions in compound word formation across languages. The objective was to determine whether large language models trained predominantly on English could effectively capture and convey the culturally nuanced cognitive strategies embedded in other languages - particularly Ukrainian, Russian, and Polish.\u003c/p\u003e\n\u003cp\u003eAn experimental setup was created using publicly accessible AI tools that simulate multilingual capabilities. The same prompt - \u0026lsquo;culturological differences in English and Ukrainian compound word formation\u0026rsquo; was submitted in four languages: English, Ukrainian, Russian, and Polish. The goal was to observe what kind of information the AI provided in each language and whether the outputs mirrored the contrastive tendencies identified in human-generated linguistic data.\u003c/p\u003e\n\u003cp\u003eThe prompt was intentionally framed in a high-level analytical style to encourage AI to generate content beyond simple definitions or grammatical explanations. The outputs were then qualitatively analyzed for thematic depth, emotional and metaphorical framing, structural complexity, and cultural relevance.\u003c/p\u003e\n\u003cp\u003eKey observations included:\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eRussian AI Output\u003c/strong\u003e: Reflected a nuanced understanding of linguistic worldview concepts. It captured\u0026nbsp;\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;emotional tone, cultural embeddedness, and metaphorical layering similar to the author\u0026rsquo;s own findings. This was likely due to greater availability of training data on this topic in Russian academic and cultural domains.\u003cbr\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eUkrainian and Polish Outputs\u003c/strong\u003e: These were largely similar to each other, often mirroring each other\u0026apos;s phrasing and structural content. While they acknowledged emotional and metaphorical differences between languages, they tended to reduce Ukrainian linguistic creativity to folk or rural imagery and\u0026nbsp;\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;emphasized agriculture-related vocabulary. The outputs appeared limited by a lack of rich training data and a tendency to replicate formal or encyclopedic language.\u003cbr\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eEnglish AI Output\u003c/strong\u003e: Focused on technical \u0026nbsp; \u0026nbsp;parameters such as structural types, productivity, joining elements, and frequency. It introduced novel categories like \u0026ldquo;tolerance for \u0026nbsp;linguistic innovation\u0026rdquo; and \u0026ldquo;influence of ideology and politics\u0026rdquo; but failed to convey emotional or metaphorical subtleties. The result was a pragmatic, stylistically neutral treatment of a topic that is inherently culturally charged.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eThe comparative AI outputs revealed a consistent pattern: emotional depth, metaphor, and cultural humor were most likely to be omitted in English outputs, whereas Slavic-language outputs - especially in Russian - reflected richer semantic and cultural nuance. These differences underline a core limitation of current multilingual AI systems: while they may translate structure or meaning, they often fail to preserve the cultural worldview encoded in linguistic form.\u003c/p\u003e\n\u003cp\u003eThis finding supports the broader hypothesis that AI, when not trained on culturally diverse datasets, reflects the biases and limitations of the dominant language paradigms it is built upon. It raises critical questions about the ethics of AI language modeling, especially as such tools increasingly shape how knowledge is generated, translated, and consumed worldwide.\u003c/p\u003e"},{"header":"4. Findings","content":"\u003ch2\u003e4.1 English vs. Ukrainian Compound Structures\u003c/h2\u003e\n\u003cp\u003eThe contrastive analysis of word formation (suffixation and compounding) in English and Ukrainian reveals fundamental differences in how each language encodes perception, emotion, and social values. These differences are not merely linguistic but cultural and cognitive, illustrating distinct worldviews.\u003c/p\u003e\n\u003cp\u003eUkrainian (as a synthetic language) \u0026nbsp;turned out to be significantly less saturated with compound new words than English (as an analytic language), because \u0026nbsp;suffixation is paramount in its derivational system.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEmotional and Diminutive Derivation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe most striking divergences in word formation in English and Ukrainian lie in the emotional expressiveness. \u0026nbsp;Ukrainian, like other Slavic languages, makes extensive use of diminutive-hypocoristic forms. These are formed not only from nouns but also from adjectives, verbs, and adverbs, using a variety of suffixes to express affection, tenderness, irony, or sarcasm. Examples: \u0026nbsp;\u003cem\u003eІваночко\u003c/em\u003e (dear Ivan), \u003cem\u003eчервоненький\u003c/em\u003e (reddish with affection), \u003cem\u003eшвиденько\u003c/em\u003e (quickly, softly), and \u003cem\u003eїстоньки\u003c/em\u003e (to eat, tenderly). English, by contrast, has a far more limited emotional derivational system, typically confined to select noun forms (e.g., \u003cem\u003edoggie\u003c/em\u003e, \u003cem\u003educkling\u003c/em\u003e, \u003cem\u003edarling\u003c/em\u003e) and lacking morphological tools to create emotional forms from verbs or adjectives.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eNature and Environment Vocabulary\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAs for the semantic groups of compounds distinguished, floral compound words (from the first semantic group) \u0026nbsp;turned out to reflect the most in common as for associative connections between the definitum and definition for English and Ukrainian word-making. Among the reasons we can outline is the fact that a significant number of names for elements of the surrounding world were inherited from classical languages. Another reason for that was the common \u0026nbsp;European environment and easy migration for cultures and languages on that territory, which caused similar associations to be evoked in representatives of two linguistic communities. For example:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ecrane\u0026rsquo;s-bill \u0026ndash; \u0026ldquo;журавець\u0026rdquo; (\u0026ldquo;zhuravets\u0026rdquo;, literally \u0026quot;crane\u0026quot; +suffix), cotton-grass \u0026ndash; \u0026ldquo;пухівка\u0026rdquo; (\u0026ldquo;pukhivka\u0026rdquo;, literally \u0026quot;fluff +suffix\u0026quot;), dog-berry - \u0026ldquo;вовчі ягоди\u0026rdquo; (\u0026ldquo;vovchi yagody\u0026rdquo;, literally \u0026quot;wolf berries\u0026quot;),finger-flower - \u0026ldquo;наперстянка\u0026rdquo; (\u0026ldquo;naperstianka\u0026rdquo;, literally \u0026quot;thimble flower\u0026quot;), etc.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;Similarities in associations underlying\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003ethe formation of names were also found in animal names.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAnthropocentric Compounds\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAnalysis of attributive composites (with attributive relationships between components) showed that English words are characterized by a significant prevalence of tactile associations. Most English composites denoting tactile associations with named objects correspond to Ukrainian composites or phrases with attributes indicating size, shape, color, aesthetic evaluation, for example, words with the following first components:\u003c/p\u003e\n\u003cp\u003e\u003cem\u003ehard\u003c/em\u003e-: -faced - похмурий, -featured \u0026ldquo;з різкими рисами\u0026rdquo;, -fisted - скупий, - grained - крупнозернистий, -nosed \u0026ldquo;тверезий, практичний\u0026rdquo;;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eheavy\u003c/em\u003e-: -browed \u0026ldquo;1. з похмурим обличчям, 2. з густими навислими бровами\u0026rdquo;, -handed \u0026ldquo;незграбний, невправний\u0026rdquo;, -headed \u0026ldquo;1. сонний, в\u0026rsquo;ялий, 2.похмурий ;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003elight\u003c/em\u003e-: -fingеred \u0026ldquo;злодійкуватий\u0026rdquo; (it is interesting, that these equivalent have different shades in connotations - positive in English, and negative in Ukrainian), -footed \u0026ldquo;прудкий, швидконогий\u0026rdquo;, -handed \u0026ldquo;спритний, умілий\u0026rdquo;, -heeled \u0026ldquo;швидконогий\u0026rdquo;;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cem\u003ecrack\u003c/em\u003e-headed (-brainеd) \u0026ldquo;недоумкуватий\u0026rdquo;, \u003cem\u003ebroken\u003c/em\u003e-bellied - \u0026ldquo;що хворіє на грижу\u0026rdquo;, etc.\u003c/p\u003e\n\u003cp\u003eSpecial attention in the study was devoted to English compound words with the components -\u003cem\u003eheart\u003c/em\u003e- and -\u003cem\u003emind\u003c/em\u003e-, which in Ukrainian words or phrases are corresponded to by formations with the lexical unit -\u003cem\u003eдуш\u003c/em\u003e- (-\u003cem\u003edush\u003c/em\u003e- literally \u0026ldquo;\u003cem\u003esoul\u003c/em\u003e\u0026rdquo;) and its synonyms, for example, words with the element -heart-:\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;heart-strings - \u0026ldquo;тайники душі\u0026rdquo;, black-hearted \u0026ndash; \u0026ldquo;з чорною душею\u0026rdquo;, heart-blood - \u0026ldquo;1. життєва енергія, 2. найдорожче\u0026rdquo;, large-hearted - \u0026ldquo;великодушний\u0026rdquo;, marble-hearted, iron-hearted - \u0026ldquo;бездушний\u0026rdquo;, down-hearted \u0026nbsp; - \u0026ldquo;занепалий духом\u0026rdquo;, open-hearted - \u0026ldquo;1) щиросердечний, 2) з відкритою душею\u0026rdquo;, plane-hearted - \u0026ldquo;простодушний\u0026rdquo;, heartsease (heart\u0026rsquo;s-ease) \u0026ndash; \u0026ldquo;душевний спокій\u0026rdquo;, heart-sick \u0026ndash; \u0026ldquo;занепалий духом\u0026rdquo;, heart-struck \u0026ndash; \u0026ldquo;зворушений до глибини душі\u0026rdquo;, false-hearted \u0026ndash; \u0026ldquo;двоєдушний\u0026rdquo;, simple-hearted \u0026ndash; \u0026ldquo;простодушний\u0026rdquo;, single-hearted \u0026ndash; \u0026ldquo;прямодушний\u0026rdquo;, hen-hearted \u0026ndash; \u0026ldquo;легкодухий\u0026rdquo; \u0026nbsp; тощо,\u003c/p\u003e\n\u003cp\u003eWith -\u003cem\u003emind\u003c/em\u003e- element:\u003c/p\u003e\n\u003cp\u003enoble-minded \u0026ndash; \u0026ldquo;великодушний\u0026rdquo;, pure-minded \u0026ndash; \u0026ldquo;чистий душею (серцем)\u0026rdquo;, single-minded \u0026ndash; \u0026ldquo;прямодушний\u0026rdquo;.\u003c/p\u003e\n\u003cp\u003eBoth Ukrainian and English reflect a common understanding of the central role of the heart as a place of focus for emotion and moral qualities. However, as we can see, in the Ukrainian linguistic picture, the element -\u003cem\u003eдуш\u003c/em\u003e- (-soul-) plays a more important role than the heart, reflecting the enduring belief of Ukrainians in the existence of the soul as an immaterial component of a person.\u003c/p\u003e\n\u003cp\u003eThis suggests that English leans toward observable traits, while Ukrainian invokes internal states and moral essence.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCulturological Vocabulary (Culturemes)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u0026nbsp; Analysis of the nominative function of occasionalisms in Ukrainian and English indicated a predominantly anthropocentric nature of occasional nomination in both languages - English and Ukrainian. The most representative thematic groups of neologisms were found to be: \u0026quot;name of a person\u0026quot; (26.2% in English and 31.6% in Ukrainian). The most relevant function of the occasional word is evaluative (which are \u0026nbsp;approximately 75% of English units of this class and 81% of Ukrainian ones), with the predominance of negative evaluations over positive ones being obvious (in a proportion of approximately 4:1 for both languages)\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;Even now, in the age of \u0026nbsp;Internet and AI \u0026nbsp;homogenization, the new compounds that appear in Ukrainian and English show huge culturological differences. Many Ukrainian occasionalisms/neologistsm are related to war reality. Here are some of the most glaring examples: \u003cem\u003eрашизм\u003c/em\u003e (rashism: Russian + fascism), \u003cem\u003eПутлєрнет\u003c/em\u003e (Putin + Hitler + Internet), and \u003cem\u003eЗатридні\u003c/em\u003e (\u0026ldquo;in three days,\u0026rdquo; mocking the unrealistic timeframe of Russia\u0026rsquo;s initial invasion claims) show the language\u0026rsquo;s capacity for expressive, politically charged compounding. These creations reveal not only linguistic ingenuity but collective emotional and cultural resistance. English, by contrast, shows a generation of compounds tied to digital life and consumerism - \u003cem\u003edoomscrolling\u003c/em\u003e, \u003cem\u003eclickbait\u003c/em\u003e, \u003cem\u003eblogosphere\u003c/em\u003e - reflecting a cultural focus on information saturation and technological immersion.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eImperative Holophrastic Compounds\u003cbr\u003e\u003c/strong\u003eEven though Ukrainian turned out to be much less productive in compounding in general, the observation yielded interesting results for one group of compounds that turned out to be more productive in Ukrainian than in English. Those \u0026nbsp; are \u0026nbsp;imperative holophrasis, representing instructions for the functional use of the named object or phenomenon: for example, wash-and-wear, cash-and-carry , pack and carry, \u0026nbsp;keepsafe, gotomeeting, do-it-yourself, pay-as-you-go, pickme, etc.\u003c/p\u003e\n\u003cp\u003eHolophrases of an imperative nature \u0026nbsp;are inherently characteristic of the Ukrainian language, as evidenced by Ukrainian surnames and nicknames. They are usually built on the following model: a verb in the imperative mood followed by a noun in the nominative case, which is usually the direct object of that verb. The combination of these two elements, often unexpected and sometimes even absurd, is a testament to the unique Ukrainian humor. Just see the examples of some nicknames: Дунь-плюнь (\u0026ldquo;Dun\u0026rsquo;-plyun\u0026rsquo; - lit. \u0026ldquo;blow and spit\u0026rdquo;, the name for someone who practices witchcraft), Ломинога (lomynoga - lit. \u0026ldquo;break leg\u0026rdquo;,the nickname for a daring dancer or a strongman who was capable of knocking down any opponent with a single blow, also for someone who was clumsy and awkward), дядько-дістань-горобчика (diad\u0026rsquo;ko-dostan\u0026rsquo;-horobchyka - lit.\u0026ldquo;uncle-get-a-sparrow\u0026rdquo;, for someone who is very tall, дівчина-розплітай-коса (divchyna-rozplitai-kosa -lit. \u0026ldquo;girl-unweave-the-braid\u0026rdquo;, for a girl of loose morals as braid was the symbol of virginity), Петро-бережи-ніс ( Petro-berezhy-nis - lit. \u0026ldquo;Peter-watch-your-nose,for someone who was clumsy and awkward).\u003c/p\u003e\n\u003cp\u003eThe last names that originated from the nicknames:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eЗаплюйсвчка (Zapluisvichka - lit. \u0026ldquo;spit-candle\u0026rdquo;, for a buzz-kill),Засядьвовк (Zasiad\u0026rsquo;vowk - lit. \u0026ldquo;ambush-wolf\u0026rdquo;), Перебийнис (Perebijnic - lit. \u0026ldquo;break-nose\u0026rdquo;),Нагнибiда (Nagnybida - lit. \u0026ldquo;bend down-misfortune\u0026rdquo;), Неїжборщ (Nejizhborsch - lit. \u0026ldquo;don\u0026rsquo;t eat borsch\u0026rdquo;),Hепийпиво (Nepyjpuvo - lit. \u0026ldquo;don\u0026rsquo;t drink beer\u0026rdquo;),Неїжхлiб (Nejizhkhlib - lit. \u0026ldquo;don\u0026rsquo;t eat bread\u0026rdquo;), etc.\u003c/p\u003e\n\u003cp\u003eThese types of compounds were typical characteristics for Old Ukrainian, unlike Old English. The earliest composites of the type pickpocket (1300) \u0026nbsp; are said to arise in the Middle English period under the influence of French imperative words (where they also appeared primarily in anthroponymy), characteristic of the colloquial style of speech [3]. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eUnlike nicknames and last names reflecting Ukrainian ironic emotional attitude to characteristics people named, Ukrainian compound first names reveal magic-like attempt to program the future behaviour of the baby, and even their destiny, eg. Miroslav is derived from Pre-Slavic and is typically given to boys. Miroslav is derived from the words \u0026ldquo;mir\u0026rdquo; (lit.\u0026lsquo;peace\u0026rsquo; or \u0026lsquo;world\u0026rsquo;) and \u0026ldquo;slava\u0026rdquo; - (lit. \u0026ldquo;glory, fame\u0026rdquo;), altogether it means \u0026ldquo;the glory for the world\u0026rdquo;, or \u0026ldquo;the glory of peace\u0026rdquo;.\u003c/p\u003e\n\u003cp\u003eВолеслав (Voleslav -lit. \u0026ldquo; glory of reedom\u0026rdquo;) , Доброслав (Dobroslav - lit. \u0026ldquo;glory of the good\u0026rdquo;), Осмомисл (Osmomysl - lit. \u0026ldquo;the one who has 8 thoughts, wise\u0026rdquo;, Броніслав/Боронислав \u0026nbsp;(Bronislav/Boronislav -lit. \u0026ldquo;glory of defense), Владислав (Vladyslav / Volodyslav, meaning \u0026quot;glory of power\u0026quot; ,Володимир (Volodymyr - meaning \u0026ldquo;owning the world\u0026rdquo;, or \u0026ldquo;ruling in peace\u0026rdquo;.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eThe comprehensive approach of Ukrainian and English compound word making analysis that was performed resulted the following conclusions:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Ukrainian compounds possess greater imagery in comparison with the English language, which is related to the more flexible morphology. This allows for the creation of more emotionally colored and expressive linguistic constructions.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Different Types of Thinking: The analytical approach of the English language reflects pragmatism and a desire for accuracy and specificity, while the synthetic approach of the Ukrainian language reflects a more holistic and figurative perception of the world.\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Values and Priorities: The emphasis on nouns in the English language reflects the focus on objectivity and the material world, while the greater focus on adjectives and verbs in the Ukrainian language reflects the value of dynamics, action, and expressiveness.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Reflection of Cultural Characteristics: Differences in attitudes between compound word components reflect cultural characteristics, such as the English aim toward practicality and the Ukrainian tendency to imagery and emotionality.\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Different Ways of Categorization: Differences in the ways compound words are formed reflect different ways of categorizing and organizing information, characteristic of speakers of these languages.\u003c/p\u003e\n\u003cp\u003eAs we see, only one linguistic sphere \u0026nbsp;(compound words) studied \u0026nbsp;and contrasted between languages within one linguistic family reveals profound cognitive and cultural differences between speakers, allowing us to come closer to unveiling of real philosophical concepts.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThese distinctions reflect cognitive and cultural orientations. \u0026nbsp;English prioritizes utility, precision, and objectivity, whereas Ukrainian compounds tend to foreground emotion, humor, spirituality, and holistic perception. These distinctions offer a window into the cognitive styles favored by each linguistic community and underscore how deeply language shapes the lived experience.\u003c/p\u003e\n\u003ch2\u003e4.2 AI Language Output Divergence\u003c/h2\u003e\n\u003cp\u003eThe multilingual AI output experiment revealed striking differences in how linguistic and cultural nuances are retained - or erased - depending on the language interface used. These discrepancies underscore how AI systems reflect the structure and limitations of their training data, often prioritizing dominant language frameworks, especially English, while failing to replicate the cognitive-emotional depth found in less represented linguistic cultures.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEnglish Output: Functional but Culturally Flat.\u0026nbsp;\u003c/strong\u003eWhen prompted in English, the AI-generated content emphasized structural characteristics of compounding: grammatical formation, productivity, stylistic usage, and semantic clarity. It introduced analytical categories such as \u003cem\u003eloanword frequency\u003c/em\u003e, \u003cem\u003eprevalence of compounding\u003c/em\u003e, and \u003cem\u003elinguistic innovation tolerance\u003c/em\u003e. However, it failed to capture metaphorical, emotional, or culturally embedded features. Descriptions were technocratic, pragmatic. Emotional coloration, humor, or symbolic significance - central to the Ukrainian compounds - were omitted entirely. The English output mirrored the data-centric, utility-focused mindset of its linguistic base.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eUkrainian and Polish Outputs: Surface-Level Cultural References.\u003c/strong\u003e When prompted in Ukrainian or Polish, the AI produced outputs that shared a common structural pattern and vocabulary, suggesting mutual translation or mirroring. These responses acknowledged cultural associations with nature and traditional life (e.g., agriculture, folklore), but rarely moved beyond superficial description. Emotional nuance, irony, and metaphorical creativity were underrepresented. The outputs often felt encyclopedic rather than reflective of lived cultural experience. This likely results from the AI\u0026rsquo;s limited access to rich, diversified training data in these languages.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRussian Output: Culturally Sensitive.\u0026nbsp;\u003c/strong\u003eThe most culturally sensitive output came in response to the Russian prompt. The AI preserved categories such as emotional tone, metaphorical layering, humor, and evaluative framing. It recognized the pragmatic nature of English compounding and the emotive, holistic tendencies of Slavic compounds. The Russian-language response aligned closely with the findings of the author\u0026rsquo;s human-led contrastive analysis, indicating that Russian likely benefits from a more robust corpus of culturally relevant linguistic studies in the AI\u0026rsquo;s training data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eComparative Implications\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAcross all outputs, the AI demonstrated a preference for structural and surface-level features when operating in or translating from English. Emotional, symbolic, and culture-specific meaning was most likely to appear in Slavic-language outputs and was almost entirely absent from English ones. Even when responding in other languages, the AI exhibited patterns consistent with English-centric cognitive models - suggesting English often serves as the implicit cognitive template for multilingual output.\u003c/p\u003e\n\u003cp\u003eThese results highlight a core limitation in the current generation of multilingual AI: the uneven representation of worldviews across languages. AI is not neutral - it is a product of its data environment.\u003c/p\u003e"},{"header":"5. Discussion","content":"\u003cp\u003eThe findings of this study highlight a crucial yet under examined tension at the intersection of artificial intelligence and linguistic diversity. While AI has become a powerful tool for global communication, content generation, and knowledge dissemination, it remains heavily shaped by the linguistic and cultural assumptions embedded in its training data. As a result, AI not only processes language -it privileges certain worldviews while marginalizing others.\u003c/p\u003e \u003cp\u003eThe contrastive analysis of English and Ukrainian compound formation revealed that these languages encode reality differently. English tends toward analytical precision and utilitarian framing, while Ukrainian favors holistic, emotionally charged, and metaphorically dense expressions. These differences are more than stylistic - they reflect divergent modes of perception, categorization, and meaning-making. However, when evaluated through AI outputs, these distinctions were inconsistently preserved. AI systems trained primarily on English data tended to reproduce English-centric patterns of thought, even when generating text in other languages.\u003c/p\u003e \u003cp\u003eThis phenomenon is a form of what is already termed \u003cb\u003elinguistic imperialism via AI\u003c/b\u003e [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. When AI systems default to English as their cognitive base, other languages, particularly those with fewer digital resources, are filtered through a narrow lens. Metaphors become flattened, emotions neutralized, and culturally resonant imagery replaced with generic formulations. The more synthetic, expressive, or context-dependent a language is, the more its unique features are likely to be misrepresented or ignored altogether.\u003c/p\u003e \u003cp\u003eMoreover, this bias is not limited to translation. It appears in AI-generated content, learning applications, search results, and recommendation engines. As AI increasingly mediates our understanding of reality, its subtle cognitive filtering shapes not only how we communicate but what we imagine to be possible or real. This creates a feedback loop: the more AI is trained on dominant languages, the more those languages dominate the informational ecosystem, which accelerates the diminishing of linguistic and cultural diversity and at the same time it obstructs tools of self-reflection for dominant language culture.\u003c/p\u003e \u003cp\u003eThis raises pressing ethical concerns. Language is not merely a technical medium; it is the vessel of memory, identity, worldview, and community. The loss of linguistic nuance is a cognitive and cultural loss. For languages already endangered or digitally marginalized, the rise of AI presents both an opportunity and a threat: an opportunity to preserve and document, and a threat of homogenization if diversity is not actively supported.\u003c/p\u003e \u003cp\u003eAI needs to be conceptualized and trained in a different way. It should be a \u003cb\u003emultimind system\u003c/b\u003e, capable of accommodating and learning from a multiple of linguistic worldviews. This requires not only more diverse training data but fundamentally different design principles: AI must be taught to recognize that different languages encode different realities, and that cultural nuance is not noise, it is signal.\u003c/p\u003e \u003cp\u003eThis insight leads directly to the proposal outlined in the next section: the creation of an \u003cem\u003eAtlas of Language Worldviews\u003c/em\u003e, a platform for capturing and amplifying linguistic diversity in the age of AI.\u003c/p\u003e"},{"header":"6. Proposal: Atlas of Language Worldviews","content":"\u003ch2\u003e6.1. Concept and Applications\u003c/h2\u003e\n\u003cp\u003eIn response to the challenges identified in the previous sections, \u0026nbsp;an \u003cem\u003eAtlas of Language Worldviews\u0026nbsp;\u003c/em\u003eis suggested to be created - a multilingual, AI-assisted platform designed to preserve, compare, and promote the diverse cognitive and cultural frameworks encoded in language, as well as to train AI itself. The goal is not merely to document vocabulary or grammatical features but to map how different language communities perceive, categorize, and emotionally relate to the world.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCreating an \u0026quot;Atlas of Worldviews\u0026quot; that engages AI and linguistic data to provide culturally-enhanced services and train AI seems to have a huge potential. The following \u003cstrong\u003e\u0026nbsp;scenarios\u0026nbsp;\u003c/strong\u003ewould be possible\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e using the Language Worldview platform:\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eLanguage learning\u003c/strong\u003e with cultural context (e.g., when and why certain phrases are used),\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eCross-cultural business\u003c/strong\u003e: AI identifies cultural misunderstandings and suggests rapport-building strategies,\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eMarketing \u0026amp; media\u003c/strong\u003e: Flags culturally inappropriate content and suggests localized alternatives,\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eHistorical interpretation\u003c/strong\u003e: Presents multiple cultural views on events.\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eCommunication aid\u003c/strong\u003e: Suggests respectful phrasings and helps navigate cultural nuances,\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eSocial media analysis\u003c/strong\u003e: Flags cultural friction and identifies trends through a worldview lens.\u003c/p\u003e\n\u003cp\u003eThe \u003cstrong\u003eculturally-enhanced service\u003c/strong\u003e the Atlas platform would result with an extended contextualized analysis of the prompt. For example, with the scenario of \u0026quot;culturological differences of compound word-making in English and Ukrainian\u0026quot;, AI\u0026apos;s Role can be to\u003c/p\u003e\n\u003cp\u003e\u0026bull; \u0026nbsp;to analyze the grammatical rules and patterns of compound word formation in English and Ukrainian \u003cstrong\u003e\u003cem\u003e\u0026nbsp;(linguistic analysis)\u003c/em\u003e\u003c/strong\u003e;\u003c/p\u003e\n\u003cp\u003e\u0026bull; to access the Atlas of Worldviews to identify relevant cultural values and beliefs associated with each language/culture (\u003cstrong\u003e\u003cem\u003ecultural contexting)\u003c/em\u003e\u003c/strong\u003e;\u003c/p\u003e\n\u003cp\u003e\u0026bull; to trace the historical development of compound word formation in each language, taking into consideration any cultural or social influences (\u003cstrong\u003e\u003cem\u003ehistorical perspective)\u003c/em\u003e\u003c/strong\u003e;\u003c/p\u003e\n\u003cp\u003e\u0026bull; to examine how the different ways of compounding words reflect different cognitive styles or ways of thinking(for example, more analytical vs. more holistic, more practical vs. more emotional) (\u003cstrong\u003e\u003cem\u003ecognitive styles)\u003c/em\u003e\u003c/strong\u003e;\u003c/p\u003e\n\u003cp\u003e\u0026bull; to analyze the metaphors embedded in compound words and how they reflect cultural values and beliefs (\u003cstrong\u003e\u003cem\u003emetaphorical analysis)\u003c/em\u003e\u003c/strong\u003e;\u003c/p\u003e\n\u003cp\u003e\u0026bull; to present the information in a clear and concise way, highlighting the cultural nuances and differences, to include references to relevant scientific studies and cultural texts (\u003cstrong\u003e\u003cem\u003eholistic analysis)\u003c/em\u003e\u003c/strong\u003e. \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe output wouldn\u0026apos;t just be a linguistic comparison but a rich, contextualized analysis that reveals the underlying cultural assumptions and perspectives that shape language.\u003c/p\u003e\n\u003ch2\u003e6.2. Roadmap of Language Woldview Atlas creation\u003c/h2\u003e\n\u003cp\u003e\u0026nbsp;Let\u0026rsquo;s try to explore a practical roadmap for developing an \u003cstrong\u003eAtlas of Worldview,\u0026nbsp;\u003c/strong\u003eoutlining the key steps.\u0026nbsp;\u003c/p\u003e\n\u003ch3\u003e\u0026nbsp;6.2.1. \u0026ldquo;Worldview\u0026rdquo; Definition\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003eThe 1st step\u0026nbsp;\u003c/strong\u003ewould be giving \u003cstrong\u003eoperational definition\u0026nbsp;\u003c/strong\u003efor the \u0026quot;\u003cstrong\u003eworldview\u003c/strong\u003e\u0026quot; notion for the Atlas. It is necessary to define what constitutes a \u0026quot;worldview\u0026quot; in a way that\u0026apos;s measurable and can be mapped across cultures. This requires moving beyond general philosophical notions and identifying concrete, observable elements. Like, for example, values, morals, beliefs, social structures, cognitive styles, relationship to nature, etc.\u003c/p\u003e\n\u003cp\u003eEach of the elements would be the answer to a set of categorizing questions.\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; \u0026nbsp;\u003cstrong\u003e\u003cem\u003eValues\u003c/em\u003e\u003c/strong\u003e: \u0026nbsp;What are the core values prioritized within a culture? (for example, individualism vs. collectivism, practicality vs. theorisation, civilizedness vs. pristineness, family values, environmental stewardship) How are these values expressed in language, traditions, and institutions?;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; \u003cstrong\u003e\u003cem\u003eBeliefs about Existence\u003c/em\u003e\u003c/strong\u003e: What are the dominant beliefs about the nature of reality, the universe, and humanity\u0026apos;s place in it? (for example, religious beliefs, spiritual practices, scientific understanding);\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; \u0026nbsp;\u003cstrong\u003e\u003cem\u003eMoral Frameworks\u003c/em\u003e\u003c/strong\u003e: What is considered right and wrong? What are the ethical principles that guide behavior?;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; \u003cstrong\u003eSocial Structures\u003c/strong\u003e: How is society organized (for example, hierarchies, kinship systems, gender roles, economic systems)?;\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eCognitive Styles\u003c/strong\u003e: How do people perceive, process, and categorize information? (for example, holistic vs. analytical thinking, tolerance for ambiguity, communication styles);\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; \u003cstrong\u003eRelationship to Nature\u003c/strong\u003e: How does a culture understand and interact with the natural world? (for example, stewardship, exploitation, reverence), etc.\u003c/p\u003e\n\u003cp\u003eOnce we have a working definition of \u0026quot;worldview,\u0026quot; it would be necessary to identify \u003cstrong\u003eobservable markers\u003c/strong\u003e. These markers will be the data points the AI would use. \u0026nbsp;For example they may be:\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Metaphors, idioms, grammatical structures, common proverbs, the way time and space are conceptualized in language - \u003cstrong\u003elinguistic features\u003c/strong\u003e;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; Art, music, literature, architecture, religious texts, cinema -\u003cstrong\u003ecultural artifacts\u003c/strong\u003e;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;◦Customs, traditions,rituals, etiquette, ethics, legal systems, political structures -\u003cstrong\u003e\u0026nbsp;social practices\u003c/strong\u003e;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; \u0026nbsp;The stories a culture tells about itself, its origins, and its heroes -\u003cstrong\u003ehistorical statements;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;The dominant intellectual frameworks used to understand the world -\u003cstrong\u003escientific and philosophical knowledge\u003c/strong\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e(!) There should be some \u0026quot;\u003cstrong\u003eStandard Worldview\u0026quot; anchor\u003c/strong\u003e, like \u0026nbsp;Greenwich prime meridian for the time zones. This anchor \u0026nbsp;shouldn\u0026rsquo;t be a superior standard, but rather a reference point for comparison, like a baseline in a statistical analysis. Explicitly identifying it as a reference point allows users to see how other worldviews differ and relate to it. It acknowledges that Western perspectives have historically dominated global discourse and academic research. There is a challenge though that it could be misinterpreted as a normative standard, perpetuating Western biases. However it seems Greenwich prime meridian hasn\u0026rsquo;t had such interpretations for being a 0 longitude. Moreover, there is a huge worldview diversity within \u0026quot;Western\u0026quot; worldviews. It is not a monolithic entity. As we already noted, there are numerous variations across European countries, North America, etc.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u0026nbsp; 6.2.2. Team\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;The next step\u0026nbsp;\u003c/strong\u003ewould be \u003cstrong\u003eforming a multidisciplinary team\u003c/strong\u003e: it would be necessary to bring together experts in linguistics, anthropology, AI, computer science, ethics, and cultural studies.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e 6.2.3. \u0026nbsp; Research Plan\u0026nbsp;\u003c/h3\u003e\n\u003cp\u003eNext, a\u003cstrong\u003e\u0026nbsp;detailed research plan\u0026nbsp;\u003c/strong\u003ewould need to be developed: the specific research questions are to be outlined, as well as data sources, and methodologies that would be used.\u003c/p\u003e\n\u003ch3\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp;6.2.4. \u0026nbsp; Funding\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u0026nbsp;\u003c/strong\u003eneeds to be secured: Funding opportunities from government agencies, foundations, and private investors are to be explored.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u0026nbsp; 6.2.5. Data Sources \u0026nbsp;\u003c/h3\u003e\n\u003ch3\u003eOne of the next steps presupposes \u003cstrong\u003especifying data and \u0026nbsp;finding data sources\u003c/strong\u003e: The AI needs to be trained on a large and diverse dataset of worldview-related information. There should be different \u003cstrong\u003etypes of data\u003c/strong\u003e used:\u003c/h3\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Linguistic Data:\u003c/p\u003e\n\u003cp\u003e○ \u0026nbsp; \u0026nbsp;Massive corpora of text and speech in multiple languages;\u003c/p\u003e\n\u003cp\u003e○ \u0026nbsp; \u0026nbsp;Grammatical structures, sentence patterns, idioms, metaphors, and slang;\u003c/p\u003e\n\u003cp\u003e○ \u0026nbsp; \u0026nbsp;Etymological information tracing the historical evolution of words and concepts.\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Cultural Data:\u003c/p\u003e\n\u003cp\u003e○ \u0026nbsp; \u0026nbsp;Historical events, religious beliefs, social norms, customs, artistic expressions, folklore, and traditions (\u003cem\u003eexplicit cultural knowledge\u003c/em\u003e);\u003c/p\u003e\n\u003cp\u003e○ \u0026nbsp; \u0026nbsp;Implicit cultural knowledge: values, assumptions, beliefs embedded in stories, proverbs, rituals, and social interactions (\u003cem\u003eexplicit cultural knowledge\u003c/em\u003e);\u003c/p\u003e\n\u003cp\u003e○ \u0026nbsp; \u0026nbsp;Gestures, body language, facial expressions, and their cultural connotations (\u003cem\u003enonverbal communication data\u003c/em\u003e);\u003c/p\u003e\n\u003cp\u003e○ \u0026nbsp; \u0026nbsp;Data on social hierarchical structures and power dynamics within different cultures.\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;User interaction data (anonymized and aggregated):\u003c/p\u003e\n\u003cp\u003e○ \u0026nbsp; \u0026nbsp;Data on how users from different cultural backgrounds interact with the platform and its associated AI services.\u003c/p\u003e\n\u003cp\u003e○ \u0026nbsp; \u0026nbsp;Feedback on the accuracy and relevance of the platform\u0026apos;s cultural insights.\u003c/p\u003e\n\u003cp\u003eWhen we think of\u003cstrong\u003e\u0026nbsp;data sources\u003c/strong\u003e:\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; A natural starting point would be UNESCO\u0026apos;s \u003cem\u003eWorld Atlas of Languages\u003c/em\u003e, It could help to explore if language families could be used as proxies for worldview similarities;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; \u003cem\u003eEthnographic databases\u003c/em\u003e could be another resource, for example., databases like the Ethnographic Atlas, the Human Relations Area Files (HRAF), and similar resources, which contain a huge amount of information about cultural practices;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; \u0026nbsp;Large collections of text and speech data (\u003cem\u003elinguistic corpora\u003c/em\u003e) can be analyzed for patterns and insights into language use;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; \u0026nbsp;\u003cem\u003eAcademic research data\u003c/em\u003e could be used to do research in anthropology, sociology, linguistics, psychology, philosophy, religious studies, and other relevant fields;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; \u0026nbsp;\u003cem\u003eOpen data sets related to cultural indicators\u003c/em\u003e, social statistics, and global values surveys could be another resource;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; \u0026nbsp;\u003cem\u003eCrowdsourcing\u003c/em\u003e, an ethically-sourced and professionally validated community, can contribute \u0026nbsp;to capturing nuances not found in formal datasets (for example, local proverbs, stories, customs).\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;(!) It is necessary to be aware of \u003cstrong\u003epotential biases\u003c/strong\u003e in existing data. Many datasets over-represent Western perspectives or focus on easily accessible cultures. It makes sense to actively seek out data from underrepresented regions and perspectives and critically evaluate the sources and methodologies used to collect the data.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Data Standardization is necessary to develop a system for coding and standardizing data from diverse sources so that it can be easily analyzed by the AI.\u003c/p\u003e\n\u003ch3\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e \u003cstrong\u003e6.2.6\u003c/strong\u003e. Platform Architecture\u003c/h3\u003e\n\u003ch3\u003e\u003cstrong\u003eThe next step\u003c/strong\u003e would be Development of the AI Platform itself.\u0026nbsp;\u003c/h3\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; It is necessary to choose an appropriate \u003cstrong\u003e\u003cem\u003eAI architecture\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e. Possibilities include: \u0026nbsp;\u003cstrong\u003eNLP \u0026amp; ML\u003c/strong\u003e - to analyze and learn cultural data ;\u003cstrong\u003e\u0026nbsp;deep learning\u003c/strong\u003e - to infer complex worldview links; \u003cstrong\u003eknowledge graphs\u003c/strong\u003e - for relational worldview mapping; \u003cstrong\u003eworldview modeling algorithms\u003c/strong\u003e\u0026nbsp; -for core belief simulation.\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; It\u0026apos;s crucial to make the AI\u0026apos;s reasoning transparent. Users need to understand the reason the AI is providing a particular response. \u0026nbsp;There should be methods developed for explaining the AI\u0026apos;s decision-making process;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; Ethics: The AI should be designed to be fair, unbiased, and respectful of cultural differences. It should not utilize stereotypes or promote certain ideologies. The AI needs to be highly sensitive to \u0026nbsp;cultural nuances and avoid making generalizations or assumptions. It should ensure that the AI is trained on data that accurately represents the diversity of worldviews. When using crowdsourced data, informed consent from contributors should be obtained and their privacy should be protected. The fact of technology misuse should be considered, and there should be safeguards implemented to prevent it.\u003c/p\u003e\n\u003ch3\u003e\u0026nbsp; \u0026nbsp; \u0026nbsp;6.2.7. \u0026nbsp; Interface\u0026nbsp;\u003c/h3\u003e\n\u003ch3\u003e\u003cstrong\u003eUser-friendly interface (\u003c/strong\u003eboth - AI training and end-user ones\u003cstrong\u003e)\u0026nbsp;\u003c/strong\u003e should be created, it should allow users to easily access and explore the Atlas of Worldviews.\u0026nbsp;\u003c/h3\u003e\n\u003cp\u003e\u003cstrong\u003eAI Training Interface\u0026nbsp;\u003c/strong\u003eshould include;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Tools for data scientists and cultural experts to annotate and validate the platform\u0026apos;s knowledge;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Visualization tools for exploring cultural differences and similarities;\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Debugging tools for identifying and correcting biases in the AI models.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEnd-User Interface\u0026nbsp;\u003c/strong\u003eis to be featured with\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Culturally-adapted output: The AI returns responses tailored to the user\u0026rsquo;s perceived cultural background, using appropriate language (\u003cstrong\u003e\u003cem\u003eculturally sensitive translation capability)\u003c/em\u003e\u003c/strong\u003e, tone, and communication style.\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;For more advanced users, the platform can provide a \u003cstrong\u003e\u003cem\u003e\u0026ldquo;cultural context\u003c/em\u003e\u003c/strong\u003e\u0026rdquo; panel alongside the AI\u0026apos;s output. This panel might include:\u003c/p\u003e\n\u003cp\u003e○ \u0026nbsp; \u0026nbsp;Explanations of culturally specific terms or concepts used in the AI\u0026apos;s response,\u003c/p\u003e\n\u003cp\u003e○ \u0026nbsp; \u0026nbsp;Information about potential cultural misunderstandings that could arise from the AI\u0026apos;s output,\u003c/p\u003e\n\u003cp\u003e○ \u0026nbsp; \u0026nbsp;Alternative phrasings or communication strategies that might be more effective in certain cultural contexts,\u003c/p\u003e\n\u003cp\u003e○ \u0026nbsp; \u0026nbsp;Relevant historical or social background information.\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp;Ways are to be suggested to rephrase prompts to be more culturally sensitive or to elicit more relevant information from the AI (\u003cstrong\u003e\u003cem\u003eprompt enhancement suggestions\u003c/em\u003e\u003c/strong\u003e).\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003e\u003cem\u003eVisualization tools\u003c/em\u003e\u003c/strong\u003e, like maps, charts, and other visual aids could help users understand complex data.\u003c/p\u003e\n\u003cp\u003e● \u0026nbsp; \u0026nbsp; Users could be allowed to explore the potential consequences of different cultural approaches to known values and beliefs (\u003cstrong\u003e\u003cem\u003einteractive simulations\u003c/em\u003e\u003c/strong\u003e).\u003c/p\u003e\n\u003ch2\u003e6.3. Language Worldview Atlas Training AI on Cultural Nuances\u003c/h2\u003e\n\u003ch3\u003e6.3.1. Data Structuring for AI Consumption\u003c/h3\u003e\n\u003cp\u003eThe Atlas would collect and organize linguistic and cultural data into structured formats:\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eLexical data\u003c/strong\u003e (e.g., culturally specific metaphors, idioms, compounds, neologisms)\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eCognitive patterns\u003c/strong\u003e (e.g., how time, space, emotion are perceived in a culture)\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003ePragmatic conventions\u003c/strong\u003e (e.g., politeness, indirectness, use of diminutives)\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eMoral and value frameworks\u003c/strong\u003e (e.g., collectivism vs. individualism)\u003c/p\u003e\n\u003cp\u003eThese would be annotated and tagged (e.g., \u0026quot;Ukrainian_holophrastic_compounds,\u0026quot; \u0026quot;English_efficiency_bias,\u0026quot; \u0026quot;Mandarin_vertical_time\u0026quot;).\u003c/p\u003e\n\u003ch3\u003e6.3.2. Integrating with Language Models\u003c/h3\u003e\n\u003cp\u003eThe structured data can then be used to \u003cstrong\u003efine-tune existing LLMs\u003c/strong\u003e or to train culturally aware models from scratch:\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eFine-tuning:\u003c/strong\u003e Inject worldview-tagged texts into models like GPT, LLaMA, or Mistral to increase sensitivity to specific cultural patterns.\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eControl tokens:\u003c/strong\u003e Embed \u0026ldquo;cultural worldview prompts\u0026rdquo; to guide generation\u0026mdash;e.g., \u0026lt;\u0026lt;Ukrainian_figurative\u0026gt;\u0026gt;, \u0026lt;\u0026lt;English_pragmatic\u0026gt;\u0026gt;.\u003c/p\u003e\n\u003cp\u003e● \u003cstrong\u003eRetrieval-Augmented Generation (RAG):\u003c/strong\u003e Use the Atlas as a semantic knowledge base to supplement AI responses dynamically during generation.\u003c/p\u003e\n\u003ch3\u003e6.3.3. Cross-Cultural Evaluation Tasks\u003c/h3\u003e\n\u003cp\u003eThe Atlas would support the creation of benchmark tasks that test:\u003c/p\u003e\n\u003cp\u003e● Can the AI generate culturally nuanced explanations?\u003c/p\u003e\n\u003cp\u003e● Can it switch metaphors appropriately across languages?\u003c/p\u003e\n\u003cp\u003e● Can it recognize worldview conflicts in multilingual input?\u003c/p\u003e\n\u003cp\u003eThis allows \u003cstrong\u003eevaluation and feedback loops\u003c/strong\u003e, improving the model\u0026apos;s ability to emulate culturally embedded reasoning.\u003c/p\u003e\n\u003ch3\u003e6.3.4. Bias Detection \u0026amp; Debiasing\u003c/h3\u003e\n\u003cp\u003eAI systems often reflect \u003cstrong\u003eWestern-centric biases\u003c/strong\u003e in:\u003c/p\u003e\n\u003cp\u003e● Moral framing\u003c/p\u003e\n\u003cp\u003e● Emotional tone\u003c/p\u003e\n\u003cp\u003e● Examples and assumptions\u003c/p\u003e\n\u003cp\u003eThe Atlas can help \u003cstrong\u003eaudit AI outputs\u003c/strong\u003e and reveal worldview imbalances by:\u003c/p\u003e\n\u003cp\u003e● Comparing outputs across worldview categories\u003c/p\u003e\n\u003cp\u003e● Flagging culturally inappropriate translations or generalizations\u003c/p\u003e\n\u003ch3\u003e6.3.5. Training AI for Intercultural Mediation\u003c/h3\u003e\n\u003cp\u003eThe Atlas could train AI assistants to:\u003c/p\u003e\n\u003cp\u003e● Detect worldview mismatches in cross-cultural communication\u003c/p\u003e\n\u003cp\u003e● Suggest culturally respectful phrasings\u003c/p\u003e\n\u003cp\u003e● Offer meta-comments like: \u003cem\u003e\u0026ldquo;Note: this phrase may carry a different tone in Ukrainian than in English.\u0026rdquo;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eNo doubt, if started small, the project may be started sooner. If we begin with a pilot project focusing on a limited number of cultures and worldview elements, and continuously evaluate and refine the Atlas of Worldviews based on user feedback and research findings, we can create a valuable resource that promotes intercultural understanding and transforms the way we interact with the world.\u003c/p\u003e"},{"header":"7. Conclusion","content":"\u003cp\u003eAs artificial intelligence continues to enter communication, education, governance, and cultural production, it is essential to recognize that AI reflects the language models it is trained on - and thus the worldviews those languages encode. If AI is trained primarily on English and other dominant global languages, it will inevitably replicate the cultural and cognitive assumptions embedded within them, often at the expense of linguistic and cultural diversity.\u003c/p\u003e\n\u003cp\u003eThis paper has illustrated how even within the same language family, profound cognitive and emotional differences exist in how speakers of English and Ukrainian percept and categorize knowledge. It has further shown that current AI systems often fail to replicate these subtleties, especially when dealing with less-resourced languages. What is lost in this process is not just linguistic detail, but a deeper capacity for comparative self-understanding, intercultural empathy, and meaningful communication.\u003c/p\u003e\n\u003cp\u003eLanguage is not just a code for information - it is a lens through which reality is perceived, remembered, and shaped. When that lens is narrowed by technology, so too is our collective imagination. If we allow AI to standardize thought, we risk silencing the plurality of human experience.\u003c/p\u003e\n\u003cp\u003eThe proposed \u003cem\u003eAtlas of Language Worldviews\u003c/em\u003e offers a concrete and scalable response to this challenge. It recognizes that each language is a repository of culturally specific cognitive patterns and that AI systems must be trained to honor this diversity rather than override it. Such a platform would not only support the preservation of endangered languages but also contribute to the creation of AI that is more ethical, responsive, and genuinely global.\u003c/p\u003e\n\u003cp\u003eTo conclude, it is necessary to outline the main opportunities of the Woldview Atlas. It can become an enhanced tool for \u0026nbsp;the following:\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u0026bull; \u0026nbsp;understanding and appreciating different worldviews (promoting intercultural understanding);\u003c/p\u003e\n\u003cp\u003e\u0026bull;\u0026nbsp; facilitating more effective communication across cultures (improving communication);\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u0026bull;\u0026nbsp; providing students with a deeper understanding of cultural diversity (enhancing education);\u003c/p\u003e\n\u003cp\u003e\u0026bull;\u0026nbsp; moving beyond simple translation to culturally-aware nuanced AI assistance (revolutionizing AI services);\u003c/p\u003e\n\u003cp\u003e\u0026bull;\u0026nbsp; facilitating more effective collaboration on global challenges (supporting global collaboration);\u003c/p\u003e\n\u003cp\u003e\u0026bull;\u0026nbsp; better negotiations, stronger relationships with international partners, and more successful marketing campaigns (more effective global business);\u003c/p\u003e\n\u003cp\u003e\u0026bull; by promoting understanding and mitigating bias, the platform could contribute to creating more inclusive and equitable societies;\u003c/p\u003e\n\u003cp\u003e\u0026bull; innovation in fields relating to languages, cultural sciences, philosophy, sociology, communication and global studies.\u003c/p\u003e\n\u003cp\u003eUltimately, the project is about expanding the scope of what we, as humans, can know about ourselves, about others, and about the world we share. In a time when digital systems are reshaping reality, ensuring that cultural multiplicity survives and thrives is not an academic luxury, it is a human necessity.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eБугров ВА (1996) Проблема мови у творчості пізнього Л.Вітгенштейна. Язык и культура: 4-я Междунар. конф.: Материалы., vol. 1, pp. 268\u0026ndash;277\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eВинокур ГО (1959) Заметки по русскому словообразованию. Избранные работы по русскому языку ed., Москв\u0026amp;#1072\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eОмельченко ЛФ (1981) Продуктивные типы сложных слов в современном английском языке (на материале прилагательных и глаголов. Киев, Вища школа. Головное изд-во\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eСеребренников БА, Уфимцева AA (eds) (1977) Языковая номинация: Общие вопросы. Москва, Наука\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAllard-Kropp (2020) Manon. Languages and Worldview. St. Louis, University of Missouri\u0026ndash;St. Louis, Open Educational Resources (OERs). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://irl.umsl.edu/oer/17\u003c/span\u003e\u003cspan address=\"https://irl.umsl.edu/oer/17\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 25 march 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eArun C (2024) Transnational AI and Corporate Imperialism. Carnegie Endowment for International Peace, https://carnegieendowment.org/, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://carnegieendowment.org/research/2024/10/transnational-ai-and-corporate-imperialism?lang=en\u003c/span\u003e\u003cspan address=\"https://carnegieendowment.org/research/2024/10/transnational-ai-and-corporate-imperialism?lang=en\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoroditsky (2019) Lera, director. How language shapes the way we think. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e\u003c/span\u003e\u003cspan address=\"http://www.you-tube.com\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, https://www.youtube.com/watch?v=RKK7wGAYP6k. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoroditsky (2021) Lera, director. How Language Shapes Thought. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e\u003c/span\u003e\u003cspan address=\"http://www.you-tube.com\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, https://www.youtube.com/watch?v=I64RtGofPW8. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoroditsky L et al (2011) Do English and Mandarin speakers think about time differently? Cognition, vol. 118, no. 1, pp. 123\u0026ndash;129. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.sciencedirect.com/\u003c/span\u003e\u003cspan address=\"https://www.sciencedirect.com/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.cognition.2010.09.010\u003c/span\u003e\u003cspan address=\"10.1016/j.cognition.2010.09.010\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCook VJ, Newson M (2007) Chomsky's universal grammar: an introduction. Wiley\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDavis G (2024) GPT-4o and Emotion Recognition Dramatically Changes AI Human Interaction. Intelligentliving. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.intelligentliving.co/gpt-4o-emotion-recognition-ai-human/\u003c/span\u003e\u003cspan address=\"https://www.intelligentliving.co/gpt-4o-emotion-recognition-ai-human/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. https://www.intelligentliving.co/, https://www.intelligentliving.co/gpt-4o-emotion-recognition-ai-human/. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJonathan Winawer J et al (2007) Russian blues reveal effects of language on color discrimination. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.pnas.org/\u003c/span\u003e\u003cspan address=\"https://www.pnas.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, vol. 104, no. 19, PNAS. https://www.pnas.org/doi/full/10.1073/pnas.0701644104. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKelly L (2019) 50 Words For Snow. https://readable.com/blog/, 8 Feruary \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://readable.com/blog/do-inuits-really-have-50-words-for-snow/\u003c/span\u003e\u003cspan address=\"https://readable.com/blog/do-inuits-really-have-50-words-for-snow/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKornai Andr\u0026aacute;s (2013) Digital Language Death. PLOS, https://plos.org/#journals. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://journals.plos.org/plosone/article?id=10.1371/journal.pone.0077056\u003c/span\u003e\u003cspan address=\"https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0077056\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMair V (2023) Where did the PIEs come from; when was that? Language Log, no. Historical linguistics, Language and archeology, Language and genetics. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://languagelog.ldc.upenn.edu/\u003c/span\u003e\u003cspan address=\"https://languagelog.ldc.upenn.edu/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e, https://languagelog.ldc.upenn.edu/nll/?p=59908. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMalotki E (1983) Hopi time: a linguistic analysis of the temporal concepts in the Hopi language. Mouton, https://ebin.pub. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://ebin.pub/hopi-time-a-linguistic-analysis-of-the-temporal-concepts-in-the-hopi-language-9783110822816-9027933499-9789027933492.html\u003c/span\u003e\u003cspan address=\"https://ebin.pub/hopi-time-a-linguistic-analysis-of-the-temporal-concepts-in-the-hopi-language-9783110822816-9027933499-9789027933492.html\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMann H (2024) AI Homogenization Is Shaping The World. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.forbes.com\u003c/span\u003e\u003cspan address=\"https://www.forbes.com\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e/, 5 March https://www.forbes.com/sites/hamiltonmann/2024/03/05/the-ai-homogenization-is-shaping-the-world/. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMarsden H (2024) \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://theweek.com/tech/how-technology-helps-and-harms-endangered-languages\u003c/span\u003e\u003cspan address=\"https://theweek.com/tech/how-technology-helps-and-harms-endangered-languages\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. The Week. www.theweek.com, https://theweek.com/tech/how-technology-helps-and-harms-endangered-languages. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMelero M. The Future of Language (and Cultural) Diversity in the Age of AI. CLARIN Annual Conference Barcelona15 October 2024. CLARIN, Barcelona Barcelona Supercomputing Center. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.clarin.eu/sites/default/files/01_Melero_CLARIN2024_Keynote_The-Future-of-Language-Diversity-in-the-Age-of-AI.pdf\u003c/span\u003e\u003cspan address=\"https://www.clarin.eu/sites/default/files/01_Melero_CLARIN2024_Keynote_The-Future-of-Language-Diversity-in-the-Age-of-AI.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNorthernLife N (2024) \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://northernlifemagazine.co.uk/neologisms-2024\u003c/span\u003e\u003cspan address=\"https://northernlifemagazine.co.uk/neologisms-2024\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e/, 26 November 2024.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eN\u0026uacute;\u0026ntilde;ez RE, Sweetser E (2006) With the Future Behind Them: Convergent Evidence FromAymara Language and Gesture in the CrosslinguisticComparison of Spatial Construals of Time. Cognitive Science, vol. 30, no. 3, pp. 401\u0026ndash;450. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://onlinelibrary.wiley.com/doi/10.1207/s15516709cog0000_62\u003c/span\u003e\u003cspan address=\"https://onlinelibrary.wiley.com/doi/10.1207/s15516709cog0000_62\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSpier L et al (eds) (1960) Language, Culture, and Personality: Essays in Memory of Edward Sapir. Greenwood Press. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://books.google.com/books/about/_Culture_and_Personality.html?id=UbJxAAAAMAAJ\u003c/span\u003e\u003cspan address=\"https://books.google.com/books/about/_Culture_and_Personality.html?id=UbJxAAAAMAAJ\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 23 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eUNESCO Headquarters in Paris (2025) Language technologies for all - Advancing Humanism through Language Technologies. lt4all2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://lt4all2025.sciencesconf.org/\u003c/span\u003e\u003cspan address=\"https://lt4all2025.sciencesconf.org/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Accessed 25 March 2025\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWhorf BL (1956) An American Indian model of the universe. Language, thought, and reality: selected writings of Benjamin Lee Whorf, no. Language and languages, pp. 57\u0026ndash;64\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWhorf BL (1941) The Relation of Habitual Thought and Behavior to Language. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://is.muni.cz\u003c/span\u003e\u003cspan address=\"https://is.muni.cz\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e/, https://is.muni.cz/el/1423/podzim2006/SAN205/um/duranti_la_whorf.pdf. Accessed 25 march 2025\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Linguistic relativity, Language Worldview, Cultural diversity, Artificial intelligence, Compound word formation, AI homogenization, Language Worldview Atlas","lastPublishedDoi":"10.21203/rs.3.rs-6719644/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6719644/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis paper examines the intersection between artificial intelligence (AI), linguistic diversity, and cultural cognition through the lens of linguistic relativity. It argues that language is not merely a communication tool but a fundamental framework shaping human cognitive perception, memory, and social behavior. Using a contrastive analysis of word formation in English and Ukrainian, the study illustrates how different linguistic structures reflect distinct worldviews. It further examines how AI development, centered mostly around English and other dominant languages, risks reinforcing cultural and linguistic homogenization. Experimental testing demonstrates how AI-generated outputs fail to capture culturally embedded emotional and metaphorical distinctions, particularly in underrepresented languages. The paper also highlights that the loss of a language entails the loss of a unique worldview and a diminished capacity for cognitive comparison. Linguistic and cultural diversity enable self-reflection, enhance communication, reduce bias, and strengthen social justice and identity. In response, the paper proposes the creation of an \u003cem\u003eAtlas of Language Worldviews\u003c/em\u003e - an AI-enhanced platform to systematically document, preserve, and map cultural perspectives across languages as well as to train AI. The Atlas would anchor a \u0026ldquo;worldview\u0026rdquo; with measurable components and integrate linguistic, historical, and cultural data, which would offer a critical tool for supporting cultural self-reflection of language groups, intercultural understanding, preserving endangered cultures, and training the ethical development of culturally sensitive AI systems. The paper provides the roadmap for Atlas creation.\u003c/p\u003e","manuscriptTitle":"AI's Blind Spot: Language Worldviews to the Rescue","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-23 11:30:43","doi":"10.21203/rs.3.rs-6719644/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"fb25c0f8-9f2b-47dc-9301-139a7bb74f3c","owner":[],"postedDate":"May 23rd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":48881838,"name":"Artificial Intelligence and Machine Learning"}],"tags":[],"updatedAt":"2025-05-23T11:30:43+00:00","versionOfRecord":[],"versionCreatedAt":"2025-05-23 11:30:43","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6719644","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6719644","identity":"rs-6719644","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}