AI-POWERED SYSTEM FOR FILTERING RELIGIOUS EDUCATIONAL CONTENT FOR YOUTH IN THE DIGITAL SPACE

Аннотация

In the current era of widespread digitalization, young people increasingly turn to the internet to seek information about religious teachings. However, the lack of content regulation has made it easier for youth to be exposed to misinformation, extremist ideologies, or distorted interpretations. This paper proposes an AI-based system for filtering religious educational content aimed at protecting youth in the digital environment. The study explores machine learning algorithms, natural language processing (NLP) techniques, and ethical considerations necessary to ensure that the filtered content is accurate, culturally appropriate, and pedagogically sound. Keywords: Artificial intelligence, content filtering, religious education, digital safety, youth, natural language processing, cyber ethics The rapid growth of digital media has transformed the way religious knowledge is disseminated, especially among young internet users. While this trend presents an opportunity for broad access to spiritual education, it also raises concerns regarding content authenticity, ideological bias, and psychological impact. Misleading religious content—whether accidental or deliberate—can negatively influence impressionable minds. Given this challenge, there is a growing demand for intelligent systems that can automatically identify and filter online religious educational material. Traditional web filtering mechanisms are often rule-based and insufficiently https://scientific-jl.com/luch/ 4 2 8 Часть-47_ Том-1_ июнь-2025 nuanced to differentiate between authentic, scholarly religious content and distorted, harmful information. Artificial Intelligence (AI), particularly Natural Language Processing (NLP), offers a promising solution to this issue by enabling systems to understand context, intent, and semantic structure of online texts. This paper proposes the design of an AI-powered content filtering framework tailored for religious educational materials targeted at youth, ensuring both safety and educational benefit in the digital space. The uncontrolled flow of digital information presents a double-edged sword, especially in the domain of religious education. On the one hand, youth now have unprecedented access to sermons, scriptures, and commentaries across different schools of thought. On the other hand, this same access leaves them vulnerable to exposure to non-accredited, ideologically extreme, or deliberately misleading content, often masked as authentic religious material. Existing content moderation systems, such as keyword filtering and manual review, are often inadequate, as they fail to capture theological nuance, semantic intent, and emotional tone—all critical factors in religious discourse. Furthermore, many social media platforms and video-sharing sites lack regionally contextualized content governance, making young users in Muslim-majority societies like Uzbekistan or Indonesia particularly vulnerable to imported radical narratives. Artificial Intelligence (AI), particularly with recent advances in transformer-based language models and context-aware classification systems, now provides a promising solution to this issue. Through machine learning (ML) and natural language processing (NLP) techniques, AI systems can be trained to distinguish between legitimate, scholarly Islamic discourse and content that misrepresents or manipulates religious principles for ideological gain. https://scientific-jl.com/luch/ 4 2 9 Часть-47_ Том-1_ июнь-2025 This research aims to address the following core questions: How can AI systems accurately detect and classify religious educational content? What metrics and ethical frameworks are needed to balance freedom of belief with content integrity? Can AI adapt to diverse Islamic perspectives while still filtering harmful content? By focusing on the development of an AI-powered filtering system, this paper seeks to support youth protection, religious integrity, and digital literacy in an age where online content increasingly shapes belief systems and identity formation. The development of the proposed filtering system involves several key stages: 2.1 Dataset Collection A diverse corpus of religious educational texts (Quranic exegesis, Hadith literature, sermons, and modern interpretations) was collected from authenticated scholarly sources and online platforms. A secondary dataset includes texts flagged for containing distorted or extremist ideologies. 2.2 Preprocessing and Labeling Texts were cleaned, tokenized, and annotated by a panel of religious scholars and linguists. Labels were assigned into categories such as: Authentic (verified scholarly content), Moderate (general religious commentary), Problematic (misinterpretation, incitement, extremism). https://scientific-jl.com/luch/ 4 3 0 Часть-47_ Том-1_ июнь-2025 2.3 Model Architecture We employed a hybrid NLP model consisting of: BERT-based language model for contextual understanding; CNN-LSTM architecture for classification; Sentimen