The advent of AI in the classroom promises to transform our teaching learning experience. This requires AI literacy and critical thinking. The integration of AI tools like Chat-GPT in the classroom offers deeper possibilities of personalization in education than never before. Open AI’s Chat-GPT and Dall-E3 , Google’s Baard, Anthropic AI Claude -2 and AI chat-bots with their sophisticated language processing abilities are set to become a game-changers in the classrooms . They provide tailored and customised learning experiences that can cater to the strengths and weakness of each student. This shift from the tradition uniform practices in education to highly individualized and personalized learning experience is going to be almost an earth shattering change in the classroom. The traditional teacher driven classroom is fast changing into student driven classroom.
By analyzing student data and putting advance algorithms to use GPT and other Large Language Models (LLMs) can create customized learning experiences that adept to each students learning style, pace and preferences. This will lead to an engaged classroom where students are active in the knowledge production as well as consumption. The banking method of education ( Paulo Friere) is all set to go. The new dynamic classroom will require AI literate teachers as well as students. Children with special needs will certainly benefit from this changed condition. Classrooms will become more inclusive making education accessible to all.
But there are downsides. The teachers might feel ill-equipped to deal with the changed scenario and the students might transfer all cognitive work to intelligent machines and become hyper passive not wanting learn . The fact that several AI tools are commercialized, some students can be at disadvantages condition as the rich students may have greater access to AI. Thus, the inclusivity that AI can bring in the classroom may be reversed. Besides, uncritical dependence of AI can lead to disastrous learning outcomes. Moreover, artificially generated texts when submitted as requirements for assessment of a student’s learning would be unethical to the degree of plagiarism. Assessments and evaluations, therefore, become complex and difficult in this new scenario. The difficulties that emerge due to the integration of AI can be overcome to a large degree through good AI guidelines that can regulate the use of AI.
The ‘ AI culture’ is yet to grip our society but when it flourishes AI’s role in the classroom will be inevitable. But we still have to critically consider AI hallucinations, AI alignment, AI runaway , AI discrimination , AI Lock-In problem. AI hallucinations occurs when AI ‘ invent’ things while sounding authoritative. AI alignment happens when AI does something other than what it was instructed to do. AI runaway occurs when AI becomes self-governing and takes up goals that were not present in our terminal instruction. AI discrimination occurs due to skewed data in its training and can lead to biased conclusions. AI Lock-In problem occurs when AI gets stuck within a narrative and lose the whole picture. Therefore, understanding the limitations as well as ways of resourceful use of its potentials in the classroom can indeed enhance teaching learning experience. Perhaps, we need a AI-ethics that considers the limitations , risks and potentials of AI in education.
But at the core of the use of AI in the classroom lies prompt engineering. Its importance in the generation of new smart learning experiences cannot be overstated. Prompting involves strategic crafting of inputs to draw desired responses from AI systems. Prompt engineering is crucial skill that enables us to effectively communicate with the AI tools. It opens ways for tailored experience of learning. By carefully designing prompts educators/ teachers can encourage students to approach problems from several perspectives, analyze critically and develop solutions creatively. AI enables us to introduce problem-solving approach in the classroom and transform the classroom into a learning space and students are transformed into learning communities. Such a mode of education will make the learning experience enjoyable and impactful.
Educators, therefore, have the challenge to embrace prompt engineering and promote best practices that deliver good learning outcomes. Since we deal mainly with Large Language Models, two major forms of prompting have been identified: Zero-shot prompts and Few-Shot prompts. Zero-Shot prompts gives a simple prompt without much details and unspecific response is generated. It says tell me something about… Few-shot prompting technique enriches the prompt with several examples that show how the task is to be competed. Few -Shot prompts can be viewed as One-Shot prompts, Two-Shot prompts , Three-Shot prompts depending on the number of examples are added to the prompt.
Prompt engineering is an experimental science and effective modes of prompting come over practice with time. There are few modes of prompting that are found effecting on LLMs. Chain-of-Thought (CoT) prompt ask AI to explain the reasoning step by step. It is discovered that narrative role plays can also yield significantly better results. This means the prompt asks the AI to put in the shoes of certain person with a specific role . This then enables the AI to be much more specific in its answer. This mode of prompting often employs expert prompting (EP) wherein the AI assumes the role of an expert.
CoT is further improved by using Self-Consistency (SC) prompting . SC-prompting asks the AI not just to follow step by step reasoning but pursue multiple lines of reasoning. SC-promoting is pursued to minimize the risk of AI hallucination. When we are not able to give any kinds of prompt to AI to get an outcome, we hit what is commonly called prompt-wise-tabula-rasa-problem. In such cases, we can take the assistance of Automatic Prompt Engineer (APE) or that of Generated Knowledge Prompting (GKn). APE starts out with one or may instances ( of texts, images, music or anything else) that the model can work with as a goal to ask AI which prompts will work best to generate them. GKn sets up the scene in which the model can operate . It constructs the framework for the AI to act as a reference to draw information from. The most complicated method of prompting is called the Tree-of-Thought (ToT) prompting. ToT uses CoT and SC prompting and goes back and forth and converges on the best line of reasoning.
AI in the classroom is a revolution and has great promises that can transform our teaching learning experiences. To deal with AI we need to rise to a new level of cognition where we are able to recognize or cognize which prompts will give us the best results while we remain sensitive to the possible limitations of AI and sufficiently critical to discern the quality of the knowledge outcome.
