Learning Outcome on Generative AI in teaching and Learning- 2

 I have been attending online sessions organized by Generative A.I. in the Teaching & Learning Group. This blog reflects my learning and the essential points discussed in the sessions.

Also read: Learning Outcome on Generative AI in Teaching and Learning- 1

Dr. Simon Ho Wang on Harnessing Generative AI for Customised Tutoring: A Practical Workshop for Educators

Generic Chatbots (like ChatGPT/GPT-4):
Pretrained on large datasets
Provide general responses across topics
Response quality can vary
Prompt engineering helps improve accuracy and relevance
An important part of developing AI literacy

Custom Chatbots:
Designed for specific workflows and audiences
Can be trained using predefined materials (e.g., documents, chapters)
Offer more structured and guided interactions
Bot’s role and behavior can be clearly defined
Useful for focused, task-specific experiences

Introduction to Poe.com:
A platform to build and deploy custom chatbots
Paid service with options to name and personalize the bot
Allows loading of specific content and instructions (system prompts)

How Custom Chatbots Work:
Preload context-specific content
Set clear roles and responsibilities
Deploy through platforms (like Poe) or API integrations

Effective Prompting Techniques:
Welcome Prompt: Introduces the bot’s purpose
Branching: Offers user-friendly interaction paths
Sequencing: Guides conversations in a step-by-step manner

Dr. Shashi Kant Shankar on Bridging Contexts with Humans-in-the-loop: The Role of Generative AI in Democratizing Quality Education

1. Understanding TPACK in a Rapidly Evolving Technological Landscape

   • Analyze how teachers can continuously develop their Technological Pedagogical and Content Knowledge (TPACK) skills in an era of rapidly advancing technology.
   • Identify strategies to balance technocentric trends with pedagogical effectiveness.

2. Exploring Teacher Perceptions on TPACK and Generative AI in TEL

   • Investigate digitally literate Indian high school STEM teachers’ perspectives on the Strengths, Weaknesses, Opportunities, and Threats (SWOT) of Generative AI in TEL contexts.
   • Evaluate the technology, pedagogy, content, and context dimensions of TPACK, including:
     • Technology: Human augmentation techniques, infrastructure gaps, experiential learning, and ethical concerns.
     • Pedagogy: Teaching enhancement, adaptation challenges, personalized learning, and transformation pitfalls.
     • Content: Relevance enhancement, digital divide, customized learning pathways, and cultural localization risks.
     • Context: Self-learning flexibility, school infrastructure challenges, equitable remote education, and balancing tradition with technology.

3. Ethical and Responsible Use of Generative AI in Education

   • Assess insights from training programs in Tamil Nadu and Bihar, revealing that despite disparities in technological infrastructure and educational support, students can effectively adapt to advanced TEL&T tools when provided context-aware training and interventions.
   • Recognize the importance of ethical considerations, digital literacy, and responsible AI usage in diverse educational settings.

4. Future Prospects of TEL&T with Generative AI at Its Core

   • Analyze the findings from The Interplay of Learning, Analytics, and Artificial Intelligence in Education: A Vision for Hybrid Intelligence by Mutlu Cukurova, which identifies four key quadrants shaping the future of education.
   • Understand the concept of Hybrid Intelligence, which emphasizes the coexistence of human intelligence and AI in education.
   • Identify key factors influencing AI adoption in mainstream education, including policy frameworks, institutional governance, pedagogical culture, technological infrastructure, and teacher support systems.
   • Recognize the importance of teacher confidence, AI literacy, and ethical considerations in ensuring AI’s successful integration into educational ecosystems.

By achieving these learning outcomes, educators will develop a holistic understanding of the interplay between technology, pedagogy, content, and context in TEL&T, ensuring that AI-driven education remains effective, inclusive, and ethically sound.

Nirmal Patel on AI Applications in Educational Assessments, Learning Engineering, and Beyond

This video explains how AI is changing education, its challenges, and how it can be improved.

1. History of AI in Education – The video discusses early EdTech innovations, like the Plato system, one of the first computers used for teaching and learning.

2. AI’s Strengths and Weaknesses – AI is great at reasoning and creating text, images, and videos, but it struggles with tasks like understanding handwriting, generating accurate geometry figures, and verifying information.

3. Pedagogy Benchmark for AI—The Pedagogy Benchmark by AI-for-Education.org checks whether AI can pass teacher certification exams. It tests AI on teaching methods, classroom management, and educational theories to ensure it effectively supports teachers.

4. Smart Paper Technology – The video introduces Smart Paper (getsmartpaper.com), which combines handwritten notes with digital tools, making learning more interactive and engaging.

5. Making AI Better for Learning – AI needs continuous testing and improvements in education. The video highlights A/B testing, game-based learning, and behavioral modeling as ways to make AI-powered tools more effective.

6. AI and Student Learning – AI can help students, but they should always double-check AI-generated answers to ensure accuracy. AI still struggles with reading fluency and handwriting recognition.

Ryan Tannenbaum on Generative A.I. literacy for teachers: Understanding how A.I. works and what you need to know for your classroom

This presentation, 'Generative AI Literacy for Teachers', provides a detailed breakdown of how Large Language Models (LLMs) like ChatGPT function and their implications in education. It highlights key aspects such as tokenization, AI biases, hallucinations, and the evolving role of educators in AI-driven learning.

Key Takeaways:

• How LLMs Work: LLMs don’t understand languages like humans; they process and generate text using tokens (small units of words or characters). ChatGPT has a vocabulary of 200,000 tokens, influencing how it interprets and generates responses. Tokens for education 

• Tokenization Impact: Common words are represented by single tokens, while less common words are broken into multiple tokens, which can affect comprehension and accuracy, particularly across different languages.

• AI Hallucinations: Since LLMs predict the next token based on probabilities rather than understanding, they can generate false or misleading information (hallucinations)—sometimes due to the "reversibility curse" (errors caused by AI's inability to perfectly reverse or reconstruct information).

• Bias in AI Responses: The way AI generates responses is not just shaped by human bias but also by language structure biases, impacting reliability.

• Educational Implications: AI's unpredictability affects grading and assessment, requiring educators to rethink evaluation strategies. Teachers should focus on guiding students in using AI critically rather than solely relying on it for answers.