Media Concept
«Proved» is a digital health media platform founded on the principles of evidence-based medicine. We stand at the intersection of science and daily life: we do not diagnose or prescribe medication, but we explain how the human body works, how medicines and tests truly function, and why certain «home remedies» can be dangerous.
Mission: To improve the quality and longevity of people’s lives by protecting them from dangerous misconceptions and fear, achieved by translating complex medical science into understandable human language. We make evidence-based medicine accessible to everyone so people can make health decisions based on facts, not myths.
Goal: To reduce anxiety levels and increase people’s medical literacy by protecting them from dangerous myths and self-medication through the provision of clear, scientifically grounded information.Sub-goals:1. Explanation: Translating complex medical terminology, medication instructions, and scientific studies into understandable, everyday language (without pretension or jargon). 2. Myth-busting: Combating popular dangerous misconceptions (from «GMOs cause infertility» to «antibiotics cure the flu»). 3. Providing tools: Teaching people how to ask the right questions of a doctor and how to understand basic health metrics (e.g., how to read a complete blood count). 4. Automating routine tasks: Using technology to monitor news, fact-check, and gather scientific data so that the editorial team can focus time on explaining meaning rather than searching articles in PubMed.
Media Themes
Segment: Health & Wellness, medical popular science, evidence-based medicine.Narrow Topic: «Evidence-Based Medicine in Daily Life» — explaining how illnesses function, how medications work, and why certain treatment methods have (or lack) scientific backing.
Content Includes:
Symptom and Disease Analysis: Why does your head hurt? How to distinguish a common cold from the flu? What is blood pressure? Pharmacology: How to take antibiotics, why you shouldn’t mix painkillers, how vitamins work. Lab Test Breakdown: What does high cholesterol mean? How to read thyroid hormone levels. Myth Busting: Checklists of popular folk remedies (from garlic in the nose to urinotherapy) — do they work or not?
Content Does NOT Include:
Providing diagnoses or prescribing treatment (we always refer to a physician). Advertising supplements, homeopathy, or dubious clinics. Clickbait headlines and manufactured panic («URGENT! THIS PRODUCT IS KILLING YOU!»).
Data Sources
International Databases of Peer-Reviewed Scientific Publications (Primary Sources)PubMed / MEDLINE (NIH) The largest database of biomedical articles. Used for searching systematic reviews and randomized controlled trials (RCTs).Cochrane Library The «gold standard» of evidence-based medicine. The database contains systematic reviews, meta-analyses, and treatment protocols with the highest level of evidence. eLibrary.ru (RINC) For monitoring relevant domestic clinical research and adapting international data to Russian realities.Official Regulators and Clinical Guidelines Sources where approved standards of medical care and official statistics are published.WHO and CDC Official statistics on morbidity, global recommendations for prevention and vaccination, and reports on disease outbreaks.Ministry of Health of the Russian Federation and Roszdravnadzor Official clinical guidelines, lists of vital pharmaceuticals, and databases on drug recalls and adverse effects (Pharmacovigilance).FDA (USA) and EMA (Europe) Primary sources regarding the registration, approval, and withdrawal of medicinal products from the market. They allow tracking global trends in pharmacology.
Structured Databases and Reference Materials (APIs and Datasets) Sources for content automation and rapid fact-checking.RxNorm / DrugBank: Databases containing information on pharmaceuticals, active ingredients, dosages, and interactions. (Used for generating «What is this pill?» cards).MedlinePlus: Informational portal of the U.S. National Library of Medicine, featuring ready-made, adapted patient guides (in English).Weather and Environmental Service APIs: (e.g., OpenWeather, air quality monitoring data) — for correlating respiratory illnesses with seasonality and environmental conditions.Official Press Releases and News Feeds (News & PR) Sources for prompt response to trending topics.RSS feeds and press offices of major medical centers and universities (e.g., Mayo Clinic, Harvard Medical School, Sechenov University). These allow us to receive news about «breakthroughs» before they are distorted by mass media.RSS feeds from regulators (Ministry of Health, Roszdravnadzor) — for instant notification to the audience about important rule changes or drug recalls.Crowdsourcing and Audience Data (User-Generated Data) Feedback forms and reader questions: Analyzing the most frequent audience inquiries (e.g., «How do I read a complete blood count?») allows us to shape the content plan based on the audience’s real needs (Data-driven content).
Data scheme separating human and machine processing
The data processing architecture in the «Proven» media platform is based on a Human-in-the-Loop (HITL) hybrid model. This approach helps mitigate the risks associated with generative models and ensures a high degree of verification for medical information. Automated systems provide scalability and speed for collecting primary data, while the cognitive functions of medical experts are directed toward critical analysis, fact-checking, and content semantic adaptation.
Phase 1. Ingestion and Aggregation of Primary Data This phase is executed exclusively through machine processing. Unstructured information is automatically collected from validated sources (PubMed, Cochrane Library, regulatory RSS feeds) using API connections and web scrapers. Raw data is placed into a repository with mandatory metadata preservation (DOI, publication date, URL), which ensures the traceability of the information in the future.Phase 2. Data Normalization and Cleaning. This process is also fully automated. Algorithms perform array deduplication, filter irrelevant traffic (e.g., predatory journals), and machine-translate specialized terminology using domain-specific medical lexicons. The result is a structured data layer, standardized into a single format and ready for semantic analysis.Phase 3. Validation and Initial Fact-Checking. Collaborative work (M+H) is initiated here. Algorithmic checks identify formal contradictions, track data currency, and flag content exhibiting clickbait or prohibited advertising traits. Subsequently, a medical editor joins the process to assess the level of evidence from the original source (for example, differentiating Randomized Controlled Trials from case reports) and decides on the appropriateness of further working with the news item. A trusted data layer is formed.
Stage 4. Semantic Generation and Editorial Adaptation This is the collaborative stage between humans and generative models (LLMs). The neural network creates a draft text based on a defined template and generates visualizations. The editor’s role is to perform semantic adaptation (translating from academic language to the target audience’s language), eliminate bureaucratic jargon, and integrate disclaimers that preclude definitive medical diagnoses.Stage 5. Expert Review This is an exclusively human quality control stage. An expert physician performs final fact verification, checks the accuracy of pharmacological data, and ensures the text aligns with current clinical guidelines. The material is marked «Verified by a Physician, ” which is a key element in building audience trust.Stage 6. Publication and Distribution This is a collaborative stage. The machine handles the technical preparation (SEO optimization, automatic tagging, and scheduling publication based on audience peak activity data). The human (the publishing editor) gives final approval for publication.Stage 7. Monitoring and Feedback During the post-publication support stage, the automation collects engagement metrics and performs a semantic analysis of user comments. The system generates alerts if new scientific data emerges (such as drug recalls) that require material updates. The human editor oversees communication with the audience and initiates the content update process, closing the feedback loop and feeding data back into the validation stage.
Guidelines for Automated Services
The «Human-in-the-Loop» Principle The key rule: the final decision regarding the publication of medical materials always remains with a human. Automated systems perform only preparatory tasks, such as information gathering, translation, and generating drafts and visualizations. The «Publish» button is exclusively in the hands of a physician-editor.Guarantee of Fact and Source Accuracy The neural network operates exclusively with verified scientific articles from our database (RAG mode), excluding the generation of information «from guesswork.» Every automatically generated piece of content includes a mandatory link to the original source (e.g., PubMed, WHO, clinical guidelines). Materials without confirmed sources or those citing sources outside the «whitelist» are blocked until manual review.Safeguards to Prevent Harm Built-in filters using stop words prevent the dissemination of dangerous content, such as promises of «100% cures,» advertising supplements, or calls to forgo medical care. Every text is automatically appended with a disclaimer reminding the reader of the necessity of consulting a doctor. Personal data from reader comments is anonymized before it is processed by the neural network.
Error Handling Procedures and Timely Data Updates In the event of technical malfunctions (such as parsing or article loading issues), the system automatically notifies the responsible editor. When official guidelines (WHO, Ministry of Health) are updated or drugs are recalled, the automation initiates a task for urgent content revision. Corrected materials are published with an «Updated» tag within 24 hours.Maintaining Calm and Preventing Panic Automated alerts regarding viral threats and disease outbreaks are generated in a neutral and calm tone. The use of trigger words (e.g., «fatal, ” „terrible“) is prohibited to prevent panic among the audience.Ensuring Algorithmic Transparency All scripts, prompts, and stop-word lists are versioned to allow for rollback of changes. A quarterly audit is conducted to check the neural network for bias and to update medical dictionaries.




