Introduction to the COSEAQ Framework

The COSEAQ framework, designed for use with generative AI, goes beyond merely facilitating educators' work with content creation. The framework aims to improve the development of high-quality educational content, such as essay questions, assessment guidelines, and quizzes, by engaging in dialogue with AI while considering theoretical aspects like constructive alignment and assessment literacy.

The COSEAQ framework is not merely a tool but rather a comprehensive scaffolding framework, designed to enhance and deepen workflows, ensuring alignment with actual teaching, the syllabus, and the national curriculum. This approach ensures that the generated content aligns with the actual teaching, syllabus, and national curriculum. The pourpose is that the framework ensure that all elements of the course work together to support student learning.

At the core of the COSEAQ framework are a series of custom GPTs (Generative Pre-trained Transformers), based on OpenAI's GPT-4. These custom GPTs build upon the concept of microprompts, which serve as scaffolding for the workflow and provide a step-by-step guide for teachers to interact with generative AI. Although the custom GPTs can function independently, they are specifically designed to be combined.

Teachers can follow the microprompts sequentially or use them as standalone prompts, depending on their preferences and requirements. This modular approach allows educators to select the GPTs that best suit their specific needs and subject matter. It is important to note that the GPTs prioritizes teacher autonomy and educational integrity and is not "an automatic machine", as this would hinder teachers' autonomy. The GPTs with the series of microprompts should be seen as scaffolding or suggestions.

By integrating generative AI into the content creation process, the framework aims to support and enhance pedagogical processes without compromising the essential role of educators. The framework reflects a combination of theoretical frameworks, allowing teachers to select approaches that best fit their teaching practice and subject matter.

Additionally, the framework can be advantageously used as separate series of microprompts in other language models, such as Claude 3 and Gemini Advanced. The complete set of microprompts for each custom GPT is available on GitHub, ensuring accessibility and flexibility for educators.

In summary, the COSEAQ framework represents a step forward in harnessing the power of generative AI for educational purposes. By providing a structured approach and a series of custom GPTs based on microprompts, the framework empowers educators to create high-quality content while enhancing their professional expertise. As the field of AI in education continues to evolve, the COSEAQ framework hopefully serves as a valuable tool for teachers to integrate generative AI into their teaching practices.

AIED #GenAI

Core Aspects of the Framework:

• **Dialogic Approach:** Facilitates meaningful exchanges between educators and AI through a series of microprompts that can be seen as the start of the dialogue, allowing for a collaborative and iterative creation process between both human expertise and GenAI.
• **Scaffolding Support:** The microprompts provide structure and guidance to generated and align the content with educational standards while promoting critical thinking and deeper understanding. These prompts are carefully designed to break down complex concepts and guide the AI in generating content that supports student learning.
• **Customized GPTs:** Each GPT is designed to stand alone, yet they can be combined to cover various educational needs, ensuring versatility and adaptability.

Purpose Beyond Efficiency: The primary goal of the COSEAQ framework is to enhance educators' professional expertise and practice, with the added benefit of streamlining their work. By integrating generative AI into the creation of educational materials, the framework supports and enhances pedagogical processes in several ways. For example, the dialogic approach encourages educators to critically examine their content and learning objectives, while the scaffolding support ensures that the generated materials align with educational standards and promote deeper understanding.

Crucially, the framework prioritizes teacher autonomy and educational integrity. Educators remain in control of the content creation process, using their professional judgment to guide the AI and make final decisions about the generated materials. The framework's flexibility, reflecting a combination of theoretical frameworks such as constructive alignment and assessment literacy, allows teachers to select approaches that best fit their teaching practice and subject matter.

Ultimately, the COSEAQ framework ensures that technology serves the teacher, not the other way around. By providing a structure for collaboration between educators and AI, the framework empowers teachers to create high-quality educational materials more efficiently while still maintaining their pedagogical expertise and autonomy.

Components overveiw

There are different Phases that have specific functions. Each Phase have specific commands that are structured set of "micro-prompts" and you move on to next by "Commands"

• COSEAQ Phase 1 - Develop Distractors and Alignment: Focus on creating plausible distractors and ensuring alignment with learning objectives.
• COSEAQ Phase 2 - Develop Study Questions: Develop questions aimed at assessing comprehension and application of knowledge for study purposes.
• COSEAQ Phase 3 - ESSAY AND TEST QUESTIONS: Formulate Assessment Instructions: Design detailed instructions for assessments to guide both educators and students.
• COSEAQ Phase 4 - QUIZ Questions Generation: Generate quiz questions that thoroughly test understanding of specific concepts, including creating distractors and providing feedback.
• COSEAQ Phase 5 - QUIZ Questions Evaluation (Beta): A beta phase focused on rigorously evaluating the generated quiz questions for plausibility, balance, and educational value.

Micropromts for each GPT

Start with Phase 1 and change between the different Phases:

Brief information / Quickstart!

Phase 1

Focuses on analyzing the concept i textbooks to pinpointing misconceptions, and aligning with the curriculum to establish clear assessment goals. The focus is on Misconception Identification, alongside ensuring Curriculum Alignment.
Upload: Textbook, curriculum and choose one or a few concepts...

Commands:

  1. READ 
    2. BREAKDOWN 
  3. Concept_Analysis

Phase 2

Design Study questions on a specific concept, ensuring depth and clarity. Include plausible distractors, grading levels, and comprehensive feedback. Step 1 to create QUIZ & ESSAY questions. See GitHub

  (DEVISE_QUESTIONS: to create questions assessing comprehension and understanding of specific concepts, including semantic intricacies.)
  1. DEVELOP_DISTRACTORS:  to generate plausible distractors based on common misunderstandings, aiming to challenge students to accurately differentiate between closely related concepts.
  3. FEEDBACK_FOR_QUESTIONS:  to provide detailed explanations for the correct answers, clarifying the reasoning behind each distractor's plausibility and addressing misconceptions.
  (FORMULATE_QUESTIONS:  to devise additional questions that deepen engagement with the subject matter, promoting critical thinking and application of knowledge.)
  2. STUDY_QUESTIONS:  to generate a set of study questions intended for revision and deeper understanding of the subject matter, exploring different aspects of the concept or area being studied.
  4. PRINT_QUESTIONS:  to compile and print all developed questions, including quiz questions, study questions, distractors, and feedback, offering a comprehensive study and revision tool.

Phase 3 BETA!

Create and refine questions on subject concepts, seeking teacher feedback. Develop assessment guidelines, balance question complexity with points, and generate examples to illustrate grading criteria. Sequentially tackle tasks, from question creation to grading facilitation.

Commands:

    STEPS_FOR_GENERATION:  to initiate the process of generating detailed assessment guidelines, including the review of policy documents, consideration of maximum points, and the balance between question complexity and points allocated.
    RESPONSE_EXAMPLES: to develop examples of student responses that illustrate different levels of understanding and meet the assessment criteria for grades E, C, and A.        
    ASSESSMENT_TABLE:  to design a table detailing each task with criteria for scoring and points allocation, facilitating assessment and grading.

Phase 4

Generate quiz questions with analytical depth, including plausible distractors and detailed feedback, focusing on a specific concept. Structure questions across difficulty levels E, C, and A, directly engaging with educational material for comprehensive assessment.

Commands:

  QUIZ_QUESTIONS: for the entire quiz question generation process, including opening the code environment, picking concepts, and creating a question table.
  DEVISE_QUESTIONS: To develop questions with depth, based on educational standards and textbook content.
  DEVELOP_DISTRACTORS: To create plausible distractors based on common misunderstandings related to the quiz questions.
  FEEDBACK_FOR_QUESTIONS: To provide explanations for the correct answers and detail why distractors are incorrect, aimed at enhancing understanding.

Phase 5 BETA!

Evaluate quiz questions through a comprehensive analysis framework focusing on plausibility of distractors, question uniqueness, answer option balance, and distractor variation to ensure educational value and effective knowledge application testing.

Commands:

  Duplicates:  to identify duplicate questions.
  Balance:  to ensure answer options are balanced and no option stands out.
  Similarities:  to find questions with high content or structural similarity.
  Plausibility:  to assess the plausibility of incorrect answers.
  Distractors:  to analyze the variety among distractors for each question.
  Report_Findings:  to compile findings and recommendations for improving quiz questions.

Theoretical approch

Comming