ACADEMIC RESEARCH IN MODERN SCIENCE
International scientific-online conference
70
CULTURALLY-AWARE AI FOR UZBEK LANGUAGE EDUCATION:
VALUES AND ORAL TRADITIONS
Boboyev Lochinbek Boymurotovich
PhD, head of IT department of Japan Digital University, lochinbek.b@jdu.uz ,
Ziyodullayev Amirbek Akmalovich
student of Japan Digital University ziyodullayevamirbek238@gmail.com
https://doi.org/10.5281/zenodo.16605828
Abstract.
This paper explores developing culturally-aware AI models for
Uzbek language education, emphasizing the integration of national values and
oral traditions. Traditional language learning often neglects crucial cultural
nuances, leading to proficient but culturally inept speakers. We argue that AI in
language education must embed cultural understanding into its core, moving
beyond homogenous datasets that risk bias. The text highlights Uzbek's rich
linguistic and cultural landscape, rooted in values like hospitality and oral
heritage (proverbs, epics). It outlines approaches for integrating these elements
into AI through cultural annotation, knowledge graphs, and interactive modules.
Challenges like data scarcity and algorithmic bias are discussed, alongside
ethical considerations and evaluation frameworks to ensure authentic, effective,
and culturally sensitive AI-powered language learning.
Key words:
AI language learning, Uzbek culture, national values, oral
traditions, cultural sensitivity
I. Introduction. In today’s increasingly interconnected world, true
communication requires more than mastering grammar and vocabulary—it
demands an in-depth appreciation of cultural norms and sensitivities.[1, 2]
Traditional language instruction, and even many AI-driven systems, often
overlook this essential cultural layer, rendering learners proficient in structure
but deficient in real-world communication skills.[1] Many such methods reduce
language learning to technical memorization, detached from the cultural
environment where language naturally operates. As a result, students often
struggle to apply their knowledge in meaningful, culturally appropriate ways.[1]
To move beyond these limitations, artificial intelligence used in language
education must treat culture not as an optional supplement but as an essential
component. Only by embedding cultural understanding into the very
architecture of AI systems can we foster learners who are not just fluent but also
capable of cross-cultural engagement.
Recent developments in AI and Natural Language Processing (NLP) have
already revolutionized education through adaptive platforms, chatbots, and AI
tutors that deliver instant feedback and personalized instruction.[2, 3] These
ACADEMIC RESEARCH IN MODERN SCIENCE
International scientific-online conference
71
tools relieve educators of repetitive tasks and help maintain student
engagement through interactive, gamified content.[2] State-of-the-art models
like GPT-3 and BERT have become central to such innovations, excelling in
language generation, contextual understanding, and translation.[1] However,
when these models are built on culturally homogenous datasets—especially
those heavily centered on Western norms—they often fail to interpret context-
sensitive expressions or reflect diverse cultural values.[1, 4] This issue,
sometimes termed “stereotype leakage,” arises when biased cultural
perspectives in training data surface in unrelated linguistic contexts, leading to
miscommunication or the unintentional reinforcement of stereotypes.[4]
Spoken by millions throughout Central Asia, Uzbek is a cornerstone of
national identity in Uzbekistan and a vessel for centuries of cultural legacy.[5, 6]
It belongs to the Karluk branch of Turkic languages and features rich dialectal
diversity—Northern variants bear traces of Russian, while Southern dialects
incorporate Persian influences.[6, 7] These dialectical differences are more than
linguistic—they reveal regional identities and historical trajectories, which in
turn shape perceptions and expressions.[7] For instance, loanwords from
Persian and Russian not only enrich the language but also reflect past political
and cultural alliances.[7]
Uzbek culture itself is rooted in time-honored values like hospitality,
reverence for elders, and a strong emphasis on family and collectivism.[5, 8, 9,
10] These principles permeate language use, from the ubiquity of kinship terms
(e.g., aka, opa, uka, singil) in social address to the customary use of indirect
praise and formal greetings.[11] Folk expressions, parables, and epics serve as
vehicles for transmitting this cultural ethos, making them essential teaching
tools.[8, 10] These oral traditions don’t just preserve stories—they encode
ethical beliefs, shared history, and linguistic subtleties that define Uzbek
identity.[10]
II. Main Part. What Is Culturally-Aware AI and Why Does It Matter in
Language Learning?
Culturally-aware AI refers to systems capable of discerning and adapting to
the cultural contexts and values of the people interacting with them.[1] Unlike
traditional AI models focused on syntactic accuracy, these systems aim for
culturally meaningful communication by recognizing idioms, social cues, and
context-specific behavior.[1, 4] For example, where Western design philosophy
might treat AI as a tool to optimize productivity, Eastern perspectives may
embrace AI as a social companion that complements human relationships.[1]
ACADEMIC RESEARCH IN MODERN SCIENCE
International scientific-online conference
72
In the language education sphere, such culturally-informed systems are
indispensable. They do more than improve fluency—they nurture intercultural
communicative competence, preparing learners to navigate real-world social
situations across cultures.[1, 12] Research demonstrates that culturally aligned
AI models outperform generic counterparts in tasks requiring politeness
detection or emotional nuance. For example, GPT-3’s accuracy in recognizing
politeness improved from 72.4% to 85.7% when enhanced with cultural data.[1]
This makes a compelling case for embedding cultural context into AI
architecture from the outset.
II. Core Principles of Culturally-Aware AI in Language Pedagogy
Understanding Cultural Sensitivity and Intelligence in AI Systems. Cultural
sensitivity in AI refers to a system’s capacity to detect and adapt to cultural cues,
ensuring respectful and accurate interaction with users from different
backgrounds.[1] This includes recognizing idiomatic language, social norms, and
context-specific meanings that go beyond surface-level translation.[1] Without
this depth, AI runs the risk of generating outputs that are not only inaccurate but
potentially offensive or misleading. Cultural Intelligence (CQ) offers a more
robust framework for building AI systems that thrive in multicultural settings.
CQ is composed of four core elements:
CQ Drive (the desire to learn about cultures),
CQ Knowledge (understanding cultural differences beyond stereotypes),
CQ Strategy (thoughtfully planning cross-cultural interactions), and
CQ Action (behavioral adaptability).[4]
Incorporating CQ into AI development helps guard against bias and
stereotyping. For instance, when English training data includes biased views—
such as negative portrayals of feminists—those views can surface in AI outputs
in entirely different languages.[4] A CQ-informed approach ensures more
diverse training data and model behavior that reflects nuanced cultural
understanding. Embedding CQ into design and evaluation practices allows
developers to build AI systems that are both fair and effective across cultural
contexts.
Opportunities and Limitations of AI/NLP in Modern Language
Learning
AI and NLP technologies have reshaped how languages are taught and
learned. These tools offer real-time feedback, personalized lesson plans, and
greater accessibility to diverse learners.[2, 3] Conversational AI and virtual
tutors simulate real-life dialogue scenarios—from ordering food to asking for
ACADEMIC RESEARCH IN MODERN SCIENCE
International scientific-online conference
73
directions—allowing students to practice in stress-free environments.[2]
Moreover, these systems adjust to individual learning styles and progress,
making instruction more effective.[2]
Advanced language models such as GPT-3 and BERT can generate fluent
text, translate between languages, and analyze context at scale.[1, 13] These
models are typically trained on vast general corpora and fine-tuned for specific
applications or languages, enabling them to work even in resource-scarce
settings.[13] However, such models often reflect cultural biases embedded in
their training data.[1] Their inability to grasp idioms, accents, or regional
dialects can undermine the communicative goals of language learning.[1] For
Uzbek learners, this means AI might correct grammar while ignoring culturally
expected forms of address or honorific speech.
Ethical concerns also arise when AI is used to preserve or teach minority
and indigenous languages. When cultural content is digitized and automated,
there’s a risk of diluting its authenticity or misrepresenting it altogether.[13, 14]
Cultural context embedded in human expression—intonation, metaphor,
ritual—is difficult for AI to replicate faithfully, and over-reliance on efficiency
could erode the richness of native communication.[13]
Theoretical Approaches for Embedding Culture in AI Education
Multiple frameworks explain how technologies like AI integrate into
education. These include:
Technology-Organization-Environment (TOE) model
Technology Acceptance Model (TAM)
Technological Pedagogical Content Knowledge (TPACK)
Socio-technical systems theory
Diffusion of Innovation theory[15]
Among these, socio-technical systems theory is especially relevant for
culturally-aware AI. It emphasizes that successful technological adoption
depends not just on tools, but on their harmony with social systems and
institutional values.[15] In other words, AI tools must align with educational
traditions, teacher practices, and community norms to be effective.
Another foundational approach is Culturally Relevant Pedagogy (CRP),
introduced by Gloria Ladson-Billings. CRP positions students' cultural
backgrounds as strengths and central to the learning process.[15] It has been
shown to enhance achievement, motivation, identity, and engagement across
disciplines.[15] By drawing on CRP principles, AI systems can be designed to
ACADEMIC RESEARCH IN MODERN SCIENCE
International scientific-online conference
74
offer localized examples, analogies, and stories that resonate with students' lived
experiences, thus supporting both educational and cultural development.[15]
Culturally relevant AI becomes a tool not just for instruction but for
inclusion—helping educators deliver lessons that are pedagogically sound and
culturally meaningful.
III. Embedding Uzbek National Values in AI-Powered Language
Learning Models
Uzbekistan’s cultural identity is deeply rooted in long-standing national
values such as hospitality, familial devotion, reverence for elders, generosity,
and a profound respect for nature. These cultural principles are not abstract—
they are intricately woven into daily social interactions and, most notably, into
the structure and usage of the Uzbek language.
One clear example is the emphasis on social hierarchy and respect, which is
linguistically represented through formal and informal pronouns. The language
differentiates between “siz” (a formal 'you') and “sen” (an informal 'you'), with
usage depending on context, relational distance, and formality. “Siz” is
customarily employed when addressing elders or individuals in formal settings,
while “sen” is reserved for close peers and family members. This distinction is a
linguistic reflection of Uzbekistan’s collective respect for hierarchy and age.
Compared to English, which generally adopts direct politeness strategies,
Uzbek relies more on subtle, indirect forms of politeness embedded in culture.
For instance, rather than a straightforward ”Can you help me?”, a polite request
in Uzbek might be framed as “Sizni bezovta qilmaymanmi? Imkoningiz bo‘lsa, bir
narsa so‘rasam maylimi?”— a phrase that conveys humility and caution in tone.
These expressions highlight the necessity of incorporating such culturally
nuanced usage into AI systems designed for Uzbek learners.
Approaches for Integrating Uzbek Values in AI Language Models:
1. Cultural Annotation During Data Preparation. Before AI models are
trained, linguistic data must be enriched with cultural annotations. This includes
tagging elements like greetings, honorifics, and formal/informal pronoun usage.
Such metadata allows AI to better understand sociolinguistic context and
generate culturally appropriate responses.
2. Data Augmentation with Cultural Context. To amplify the training dataset
and embed national values, culturally resonant alternatives should be
substituted for neutral or generic phrases. For example, replacing “How are
you?” with “Salomatmisiz?” or a culturally contextual greeting enhances cultural
ACADEMIC RESEARCH IN MODERN SCIENCE
International scientific-online conference
75
depth and realism. The augmented data should strike a balance to prevent
overfitting while maintaining diversity.
3. Use of Knowledge Graphs (KGs). Knowledge graphs provide a structured
way to map relationships between linguistic elements and cultural practices. In
pre-training, KGs can add semantic layers to raw data, linking address forms to
social norms. During training, KGs can be integrated into the model’s neural
structure, guiding it to recognize social hierarchies or hospitality values. After
training, these graphs aid in explaining AI decisions through culturally grounded
logic, improving transparency and trustworthiness.
4. Neuro-Symbolic Techniques. Combining symbolic reasoning (via KGs)
with neural models enables AI to reason through cultural associations and
linguistic patterns. For instance, a proverb may trigger a chain of cultural
inferences, allowing the AI to not only recognize it but apply it appropriately in
dialogue or storytelling
Teaching Uzbek Cultural Values via AI.
1. Interactive Cultural Modules. AI systems can deliver real-time, engaging
modules built from up-to-date Uzbek content—like news stories, interviews, and
folklore. Lessons can simulate real-life conversations where learners practice
honorific greetings or respectful behavior toward elders. The AI offers real-time
correction to ensure learners internalize both language and etiquette.
2. Adaptive Cultural Instruction. Using learner analytics, AI platforms can
deliver personalized learning pathways. Beginners might receive simplified folk
stories annotated with vocabulary support, while advanced learners engage
with complex cultural texts like proverbs or philosophical parables.
3. Personalization with Cultural Anchoring. AI systems can tailor content to
learners’ interests—offering stories, proverbs, or news based on thematic
preferences such as music, architecture, or traditions. This reinforces cultural
learning by connecting language to personal engagement.
4. Cultural Role-Play and Simulation. Large Language Models (LLMs) can
simulate interactions with virtual Uzbek locals, allowing students to test their
knowledge of social norms in safe environments. Such systems should, however,
be carefully trained to avoid reinforcing cultural stereotypes inadvertently
present in training data.
IV. Integrating Uzbek Oral Traditions into AI Language Instruction
Uzbek oral traditions—proverbs, epics, legends, folk tales—serve as the
cultural DNA of the nation. They transmit wisdom, ethical guidance, and
ACADEMIC RESEARCH IN MODERN SCIENCE
International scientific-online conference
76
communal values through vivid narratives passed down orally for generations in
families, schools, and community settings.
Proverbs such as “Yaxshilik qil, daryoqa tashla” (Do good and throw it into
the river) reflect collective wisdom and social ethics. Epics like Alpomish and
tales like Tomaris represent resilience, patriotism, and familial love, helping
shape youth identities through culturally charged storytelling.
Integrating Oral Traditions in AI Datasets
1.Digitization and Cultural Tagging Many oral traditions exist in spoken
form or scattered texts. Gathering these materials, digitizing them, and applying
cultural annotations (e.g., values, region, intended moral lesson) are crucial first
steps. Community collaboration ensures accuracy and authenticity.
2.Structuring
Folk
Knowledge
via
Knowledge
Graphs
Proverbs and tales encode societal norms. KGs can map these sayings to values
(e.g., patience, family honor) and situations (e.g., weddings, conflict resolution),
enabling AI systems to understand not just what is said, but why it is culturally
significant.
3.Augmenting Folklore Data Using generative tools, folk stories can be
adapted into modern contexts or transformed into role-play exercises. Proverbs
can be embedded into everyday dialogues for learners to practice in AI-based
tutoring systems.
4.Transfer Learning for Low-Resource Languages Pretrained multilingual
models (e.g., mBERT, XLM-R) can be fine-tuned with culturally annotated Uzbek
data. Even with limited resources, this method enables the creation of robust,
culturally fluent language models.
Teaching Oral Traditions Through AI
1.Culturally-Aware Chatbots Chatbots can challenge learners with
culturally relevant tasks—e.g., “Give advice to a friend using a proverb.” The
system evaluates not just grammar but also cultural appropriateness.
2.Dynamic Content Generation AI tools can generate age-appropriate
versions of folk stories, epics, and idioms, creating exercises that explore the
lesson or value of each piece.
3.Pronunciation and Accent Recognition Oral heritage is intimately tied to
pronunciation and dialect. AI tools should recognize and guide learners through
dialect-specific features, helping them mirror native intonation.
4.Culturally-Oriented Learning Analytics AI can track learners’ use of
idioms, proverbs, and politeness forms. Teachers can use this insight to adjust
instruction and reinforce cultural awareness.
ACADEMIC RESEARCH IN MODERN SCIENCE
International scientific-online conference
77
III Conclusion and Strategic Recommendations
.
Creating culturally-competent AI systems for Uzbek language learning
offers a transformative opportunity to preserve heritage while advancing
education. Language cannot be divorced from culture; thus, AI systems must
embed both to truly support learners.
Strategic Steps Forward: Culturally-Enriched Datasets: Collect, annotate,
and expand data rooted in Uzbek values, etiquette, and oral heritage; Knowledge
Graph Integration: Structure cultural logic into AI systems for more meaningful
language generation and comprehension; Pedagogical Innovation: Build
adaptive, culturally relevant modules and simulations using folk content and
real-world social contexts; Personalized Cultural Pathways: Use AI to deliver
content matched to learners’ interests, backgrounds, and cultural connections.
By weaving culture into the fabric of AI development, we move closer to creating
systems that are not only smart—but wise
References:
1.
Cultural Sensitivity in AI Language Learning: Using NLP to Enhance
Language Understanding Across Cultural Contexts | Theoretical and Natural
Science. https://www.ewadirect.com/proceedings/tns/article/view/20110
2.
How
AI
Improves
the
Language
Learning
Experience.
https://www.intellectsoft.net/blog/how-ai-improves-the-language-learning-
experience/
3.
Applications of NLP in Education. https://tezeract.ai/applications-of-nlp-
in-education/
4.
AI
and
Learning
About
Cultures.
https://www.centerforengagedlearning.org/ai-and-learning-about-cultures/
5.
The
expression
of
Uzbek
national
values
in
English
translation.(http://worldlyjournals.com/index.php/IJSR/article/view/8092)
6.
DIALECTOLOGY: DIALECTS OF THE UZBEK LANGUAGE AND THEIR
CHARACTERISTICS.
https://webofjournals.com/index.php/1/article/download/2615/2585/5093
7.
Uzbek
language
-
Wikipedia.
https://en.wikipedia.org/wiki/Uzbek_language
8.
Proverbs and sayings are also genres of Uzbek oral folk art.
https://ichlinks.com/archive/elements/elementsV.do?elementsUid=13829896
140737001052
9.
The development of the labor process, the experience of people, and the
development of consciousness have laid the groundwork for the development of
ACADEMIC RESEARCH IN MODERN SCIENCE
International scientific-online conference
78
oral literature. https://media.neliti.com/media/publications/428885-none-
3dca0fdd.pdf
10.
A COMPARATIVE ANALYSIS OF UZBEK AND ENGLISH PROVERBS:
CULTURAL
INSIGHTS
AND
LINGUISTIC
FEATURES.(https://www.researchgate.net/publication/389119555_A_COMPAR
ATIVE_ANALYSIS_OF_UZBEK_AND_ENGLISH_PROVERBS_CULTURAL_INSIGHTS_
AND_LINGUISTIC_FEATURES)
11.
Ethical and Privacy Considerations in AI-Driven Language Learning -
Dialnet. https://dialnet.unirioja.es/descarga/articulo/10118101.pdf
12.
AI
Ethics
And
Language
Learning.
https://www.meegle.com/en_us/topics/ai-ethics/ai-ethics-and-language-
learning
13.
A Theoretical Framework for Interrogating the Integration of Artificial
Intelligence
in
Education
-
ResearchGate.(https://www.researchgate.net/publication/384412406_A_Theor
etical_Framework_for_Interrogating_the_Integration_of_Artificial_Intelligence_in
_Education)
14.
Generative AI and Large Language Models in Language Preservation:
Opportunities and Challenges - arXiv. https://arxiv.org/html/2501.11496v1
15.
Preserving the Past: AI in Indigenous Language Preservation - Viterbi
Conversations in Ethics. https://vce.usc.edu/weekly-news-profile/preserving-
the-past-ai-in-indigenous-language-preservation/
