QuestionAid

Accepts educational content (text, documents, or course materials) and uses large language models to automatically generate assessment questions across multiple formats. The system likely employs prompt engineering or fine-tuned models to extract key concepts and generate pedagogically-structured questions with configurable difficulty levels, then structures outputs as question objects with metadata (difficulty, question type, correct answer, distractors).

Translates generated question objects into Moodle-compatible XML/GIFT format and pushes them directly into Moodle instances via API or file upload, eliminating manual import workflows. The system maintains question metadata (difficulty, tags, learning objectives) during format conversion and handles Moodle-specific constraints (question bank organization, category hierarchies, question type limitations).

Converts generated questions into Kahoot-compatible format (JSON or Kahoot API calls) with automatic adaptation for game-based learning constraints: enforces 4-option multiple choice, applies time limits, assigns point values, and structures questions for real-time classroom delivery. The system maps question difficulty to Kahoot point multipliers and handles Kahoot's specific metadata requirements (quiz name, description, cover image, player limits).

Allows educators to specify target difficulty levels (e.g., Bloom's taxonomy levels: remember, understand, apply, analyze, evaluate, create) and generates questions aligned to those cognitive levels. The system uses prompt engineering or classification models to ensure generated questions match specified difficulty, then allows post-generation adjustment of difficulty ratings before export to LMS platforms.

Supports generation of multiple question formats (multiple choice, true/false, short answer, matching) from the same source content and allows educators to specify the distribution of question types in bulk exports. The system applies format-specific generation logic: MC questions include plausible distractors, T/F questions avoid ambiguity, short answer questions define acceptable answer variations, and matching questions pair related concepts.

Processes large question batches (50-500+ questions) asynchronously with progress tracking, error reporting, and partial success handling. The system queues generation requests, monitors LLM API usage and rate limits, retries failed generations, and provides educators with real-time or post-completion reports on generation success rates, quality metrics, and any questions requiring manual review.

Analyzes generated questions against source content to detect factual errors, ambiguous distractors, and misaligned learning objectives. The system uses semantic similarity matching, fact-checking heuristics, and pedagogical rules to flag questions requiring manual review before export. Validation includes checks for: answer key correctness, distractor plausibility, question clarity, and alignment with stated learning outcomes.

Maps generated questions to specified learning objectives (e.g., BLOOM's taxonomy, state standards, course outcomes) and allows educators to filter, organize, and export questions by learning objective. The system uses semantic matching to align questions with objectives, then provides visibility into which objectives are well-covered and which need additional questions.

Organizes generated questions into hierarchical question banks with automatic tagging by topic, difficulty, question type, learning objective, and custom educator-defined tags. The system creates or updates question bank structures in both QuestionAid and target LMS platforms (Moodle, Kahoot), maintaining tag consistency across platforms and enabling educators to search, filter, and export questions by any tag combination.