AI Learning Personalization
Client
Binus University
Overview
In response to changes in the educational landscape, our company assisted BINUS University in utilizing its extensive database to create a personalized learning recommendation system. The system provides course recommendations designed to create tailored learning paths that help individuals advance in their careers or job positions based on their interests and goals. These recommendations are determined by generated competencies by the BINUS Center for Learning & Development (BCL&D), supporting all BINUS students' learning and development, for each course and job position. The system matches these competencies and aligns them with the individual's interests and goals to provide the most relevant and beneficial course suggestions.
The system integrates Generative AI to enhance competency generation from the database and suggest new competencies for continuous improvement. Previously, BCL&D could only manually generate average 10 competencies per course daily. With Generative AI assistance, all competencies for every course can now be generated with a single click.
Challenges
Needed to modernize their educational approach by utilizing an extensive database of competencies for courses and job positions.
Provide recommendations, and integrate seamlessly with their existing portal.
Solutions
Implemented an asynchronous competency generation system using AI
Designed a quality assurance process to ensure reliable AI-generated results.
Created an intuitive interface and integrated it seamlessly into the existing portal.
Developed a non-intrusive feedback mechanism to gather user insights
Decision Making
This project have heavily on the making for generate competency and approval system for BCL&D (Figure 1.0).While for interface it self there is not many things that I did since they are already have the design system.
Fig 1.0 Flow Generate Competency
Decision 1: Asynchronous Competency Generation
Utilizing generative AI to produce competencies and offer suggestions can be time-consuming. It was crucial to prevent users from being idle while waiting for this process to complete. To address this, we implemented an asynchronous system, allowing BCL&D to leave the platform during competency generation. We enhanced the user experience by informing them that the process is asynchronous and requires no further action until completion. We designed four distinct states represented as tabs for better organization: Completed, In Progress, Without Competency, and Failed. This organization helps BCL&D efficiently manage actions, such as regenerating competencies in the Completed state, generating new ones in the Without Competency state, and retrying in the Failed state.
Fig 2.0 Generate Competency
Decision 2: Quality Assurance for Generated Competencies
To ensure the quality of AI-generated competencies and prevent potential hallucinations, a rechecking process was necessary. Discussions with stakeholders revealed the need for supervisor verification and editing capabilities. Since multiple supervisors might make edits, tracking these changes was vital. We implemented a history feature similar to Google Docs and GitHub's timeline cards, displaying what changes were made, by whom, and when. This feature promotes transparency and facilitates effective collaboration during the limited approval phase.
Fig 3.0 History & Versioning
Decision 3: Seamless Integration with BINUS Portal
The recommendation system was embedded within the BINUS portal, necessitating a consistent look and feel. The project's success was measured by achieving an 80% recommendation criteria match. To evaluate this, we conducted a survey over a set period, allowing users to provide natural, unbiased feedback. This approach ensured genuine responses without pressuring users to select specific options. The most comfortable method was to create a modal that appears unobtrusively when the user opens the portal, prompting feedback after they have interacted with the recommendation system. The feedback collected through these surveys is analyzed to make data-driven improvements, ensuring continuous enhancement of the system based on user insights. Future iterations will incorporate additional features and refinements to further align with user needs and preferences.
*The interface is different from the previous version due to an update from BINUS during the project's development. Iterations were held to align it with the updated design system.
Fig 4.0 Recommendation for Interest
Fig 4.1 Recommendation for
Switch Career Paths
Fig 5.0 Design Preview
Current Status
The development phase of the project is now complete, and the stakeholders are pleased with the results and functionality. However, further iterations are still ongoing to adjust and refine the system based on stakeholder feedback and evolving needs.
As mentioned in the overview, the initial process required 50 days for one person to manually handle sample data for 25 courses and 25 positions. By utilizing generative AI, this process has been significantly accelerated, reducing the time required to less than a week. This improvement demonstrates the effectiveness and efficiency gained through AI integration.