HumanAIze hackathon submission - AI based learning companion

Combining AI and edtech has the potential to revolutionize the way we approach learning and knowledge acquisition. However, we are more limited by ineffective studying methods and cannot learn and retain information efficiently. Keeping this in mind, I had an idea I wanted to implement, and the HumanAIze hackathon on Hack2Skill seemed perfect for this. So, without further ado, here's my idea submission for the hackathon.

tags: Hack2skill, GenAI, AI, HumanAIze, Hackathon

My idea

Hack2skill, GenAI, AI, HumanAIze, HackathonA learning accountability platform with a personalised learning planner and an AI accountability and learning partner, using a chat UI. The AI does not generate content for the user to learn; instead, it motivates the user to learn better using proven learning techniques and instils a learning habit into them. It does not act as a mentor, rather it is at an equal level, as a learning partner you talk with after studying.

The flexibility of a chat based platform allows AI to interact with the user directly, enabling it to employ well known learning techniques to enhance their understanding. Based on the user's progress, it will employ techniques such as spaced repetition, inquiry based learning, the Feynman technique and more, depending on the user's preferences.

What inspired this

In the course of thinking up an idea for this hackathon, I researched many methods of studying, memorization techniques, learning patterns, goal and habit setting as well as methods to keep the mind engaged, day after day of studying.

Quite a few things stood out - setting goals, creating habits and habit stacking (atomic habits), taking notes using Zettelkasten, spaced repetition for memorization, and keeping accountability partners to improve consistency. Since the methods are so varied in nature as well, I decided to implement a flexible platform for it, where it can share notes, links to videos, lessons etc. as well as flashcards and motivate the user as well.

A chat-based platform seemed perfect for this.

...and then I hit a roadblock

Initially, I wanted to make everything a smooth experience for the user. I thought that they should have to expend no effort, and simply enjoy the fruits of their learning.

However, along the way, I realised that the person has to put in effort in order to use these methods AND actually learn. They actually have to make changes in their life This barrier of effort(and resources) often discourages people from actually using them.

But, making this easier should not mean that the user puts in less effort to actually study, as this can make the learning itself seem mindless and less fulfilling. This means that we cannot use AI generated short notes/cramming as a shortcut. Moreover, the factual correctness and grounding of AI generated content pose major concerns.

Hence, I decided to use AI to instead scour the internet for relevant resources - including textbooks such as LibreTexts, blogs and online resources to draft a learning plan - a personalised learning plan creator. An AI agent then decides which resources would be best for the way would be best to approach learning for the user and implements this. It also shares the resources relevant to the current step of the plan to the user - which can range from links to articles and videos to an AI generated summary of a research paper that we need to learn.

This data is fed to the AI as well, enabling it to converse about the topic, ask and answer questions. Hence, it can act as an accountability partner as well, even asking the user to explain what they learnt two days ago.

Potential Impact:

By providing a personalised, interactive, and adaptive learning experience, it can enhance engagement, retention, and ultimately, mastery of the subject matter. The platform's emphasis on accountability and motivation can foster consistent learning habits, leading to improved academic or professional outcomes. With its unique approach and the power of AI, this platform can significantly impact how people learn and retain information.

A general overview of how it should work

• The AI creates a learning plan, and based on that will send the user learning resources to complete each step. The user then studies the topic using the given resources, or they can even find some more.

• Then, when the user returns, the AI converses with the user, mainly about the topic. After each step, it judges the user's understanding by asking them to explain each topic.

• Based on this the user's learning plan is adjusted, and more fundamental or advanced topics are added depending upon their levels of mastery.

• Moreover, many tools and techniques such as spaced repetition, flashcards, Socratic questioning, the Feynman technique etc. can be configured by the user, and will be used by the AI when appropriate to do so.

• The AI can also motivate/suggest the user to use tools such as Zettelkasten and habit stacking.

I plan on using an agentic workflow as well as RAG to enhance the AI's knowledge and decision-making power. Each technique can be used as a tool, and thus makes the system very flexible.

Tech Stack

• React, along with Tailwind CSS for the frontend

• Python and FastAPI to manage the backend

• The AI will be managed using both self-hosted and cloud-based models using openrouter

• The AI framework of choice is LlamaIndex as I have prior experience in it, and it works well for document based workflows

• Milvus as a vector database for RAG.

Conclusion

Thank you for taking the time to read through my idea.
By combining the power of artificial intelligence with proven mind techniques, I want to create a truly personalized, engaging, and effective learning experience. I hope that you can see the potential impact it could have on students, professionals, and learners alike. I'm excited at the thought of bringing it to life!