Blog

MindGuideAI Mentor for Children (API)

Contents
  1. 1. Goals and Objectives
  2. 1.1 Main Goals
  3. 1.2 Additional Tasks
  4. 2. Key Working Principles
  5. 3. General Structure and Microservices Approach
  6. 3.1 System Modules
  7. 4. Interaction with External AI (ChatGPT.com, DeepSeek.com)
  8. 5. Logic of Operation with Related Topics
  9. 6. Mechanisms for Engaging the Child
  10. 7. Moderation and Safety
  11. 8. White Label: Licensing and Integration
  12. 9. Analytics Collection and Results Evaluation
  13. 10. Application Architecture Proposal (Tech Stack)
  14. Simplified Interaction Scheme
  15. 11. Summary and Benefits

1. Goals and Objectives

1.1 Main Goals

  • Educational: Encourage children to think critically rather than passively receive ready-made solutions.
  • Developmental: Help children develop missing or weak skills (soft/hard skills) through a system of related topics and questions.
  • Engaging: Maintain the child’s interest and motivation by providing content relevant to their preferences.

1.2 Additional Tasks

  • Skill Screening: Create a profile (competency map) reflecting the child’s strengths and weaknesses.
  • Safety and Moderation: Filter inappropriate content and protect against toxic or prohibited topics.
  • White Label: Enable integration of individual modules (dialogue filter, analytics, moderation) into third-party solutions.

2. Key Working Principles

  1. Expanding the Chain of Reasoning
    • For every child’s query, the system asks follow-up questions, encourages reflection, and provides hints instead of offering instant “ready-made answers.”
  2. Recommending Related Topics
    • Based on interest analysis (e.g., a child likes biology but also needs chemistry basics), the system suggests, “You might also find it interesting to learn how…” or “Let’s explore how these processes are similar…”.
  3. Identifying and Developing Missing Skills
    • If the system notices a child consistently struggling with arithmetic, it can subtly include exercises or questions to practice these skills within the context of the child’s interests.
  4. Learning Through Interest
    • Uses topics that already engage the child (search patterns, keywords, query analysis) and offers extended tasks to encourage deeper exploration.

3. General Structure and Microservices Approach

The project is implemented as a set of microservices, each serving a specific role. This architecture ensures flexibility, scalability, and the possibility of White Label solutions.

3.1 System Modules

  1. UI / Frontend
    • Child-friendly interface (mobile app and/or web portal).
    • Optional: Modules for parents and/or teachers with reports and settings.
    • Integration capability via iFrame/SDK or a custom web client for White Label.
  2. API Gateway / Orchestrator
    • Receives requests from the frontend.
    • Distributes them among other services (Dialogue Service, Analytics Service, Moderation Service).
  3. Dialogue Service (Dialogue Filter)
    • Core “brain” of the communication process:
      • Receives a child’s question, classifies it (type, complexity, etc.).
      • Generates an extended query to ChatGPT.com or DeepSeek.com considering the child’s age group and skill development goals.
      • Receives a response from the external service and reformats it for the child: adds guiding questions, simplifies terminology, includes visual hints, etc.
      • Tracks the dialogue chain (context across multiple messages), suggests related topics and exercises.
  4. Analytics Service
    • Stores and processes dialogue information: time, topics, child’s responses, difficulties, errors, preferences, etc.
    • Creates a skill profile (soft/hard) based on the child’s behavior (e.g., frequently asks for math hints → weak math skills; enjoys asking “why” about animals → interest in biology).
    • Provides recommendations to the Dialogue Service on which topics/skills to further develop.
  5. Moderation Service
    • Automatic moderation (AI model trained to detect toxic or inappropriate queries).
    • Manual moderation (operators who review suspicious sessions, complaints, etc.).
    • Filters unacceptable content and blocks/edits responses from external AI.
  6. Recommendation Service (Analytics Add-on)
    • Based on the child’s profile, suggests “interesting” or “useful” topics and offers expansions.
    • Considers age range, current school curricula, and trends in children’s education.
  7. Database
    • Stores dialogue sessions, child profiles, skill data, moderation logs.
    • Separately – logs of requests to ChatGPT.com / DeepSeek.com and responses.
  8. Admin Panel
    • Interface for moderation operators.
    • Settings for censorship levels, filter keywords, incident reports.
    • Panel for managing White Label clients (issuing keys, licensing individual modules, etc.).

4. Interaction with External AI (ChatGPT.com, DeepSeek.com)

  1. Connection via API: Uses official APIs (e.g., OpenAI API for ChatGPT).
  2. Dialogue Filter Layer wraps all requests and responses to:
    • Consider the child’s age and question complexity.
    • Add guiding questions before displaying the answer.
    • Filter or adapt text if it doesn’t meet the criteria (complex terminology, inappropriate content).
  3. DeepSeek.com (depending on capabilities) can supplement the knowledge base with specific data to provide more accurate answers or a wider range of content.
  4. Moderation occurs before and after querying ChatGPT.com/DeepSeek.com to:
    • Prevent sending inappropriate/dangerous queries to external services.
    • Check responses for toxicity, violations, overly mature content, etc.

5. Logic of Operation with Related Topics

Here’s an example of a step-by-step scenario:

  1. A child asks: “Why do cats purr?”
  2. Dialogue Service contacts the Analytics Service to determine:
    • The child’s age (e.g., 10 years old).
    • Current knowledge level in biology (known from previous sessions).
    • Tendency to ask “why?” questions (above average).
  3. Dialogue Service forms an extended query to ChatGPT with instructions:
    • Provide a simplified scientific explanation.
    • Include 1–2 guiding questions (e.g., “Why do you think cats purr?”).
    • Suggest a related topic about animal behavior (“Let’s think about how dogs express joy and how it’s different?”).
  4. ChatGPT returns a response (raw data), and the Moderation Service checks it for age-appropriateness and correctness.
  5. Dialogue Service structures the response into a step-by-step presentation:
    • Step 1: “Have you ever noticed that cats purr not only when petted? Why do you think that is?”
    • (The child responds.)
    • Step 2: “Actually, purring helps cats… (small hint). Do you think it’s related to their well-being?”
    • (The child tries to answer again.)
    • Step 3: Final explanation + invitation to a related topic: “Would you like to learn how other animals express emotions?”
  6. If the child agrees, the system suggests an extended topic (e.g., “How dogs, horses, or other animals express joy, fear, etc.”).
  7. Analytics Service receives dialogue logs, updates the “skill map” (the child showed great interest in biological topics). It may note that the child didn’t know the term “innate instincts” and suggest introducing other terms later.

Thus, each dialogue becomes not just an answer but an educational mini-session with expanded knowledge and skill acquisition.

6. Mechanisms for Engaging the Child

  • Gamification: Points, badges, levels like “Explorer” or “Inventor.”
  • Visual Elements: Images, emojis, mini-videos, audio accompanying explanations.
  • Simplified Analogies: For complex scientific facts, use life examples (comparisons, stories, fairy tales).
  • Rewards for Engagement: If the child shows interest in a new topic or answers additional questions, reward them (virtual stickers, certificates).