LARGE LANGUAGE MODEL (LLM)
STUDENT ASSISTANT
SEPTEMBER 2023 - PRESENT (ONGOING)
20 - PERSON TEAM - FRONT END & BACK END, PRODUCT MANAGEMENT, UI/UX, LLM
MY ROLE - UI/UX DESIGNER
THE CHALLENGE
"Will ChatGPT know that you are a student?"
Let's keep it real. The current Chatgpt lacks customization and personal modification. When you chat with Gpt, he doesn't know that you are a student, and he doesn't know what you want as a student.
Our app aims to address a critical gap in the current AI conversational landscape: the lack of personalization and customization. When engaging with AI, users often encounter a one-size-fits-all approach, which fails to recognize individual identities, such as being a student, or understand specific desires tied to those identities. This gap can make it challenging for our target users, especially those with unique needs like students, to leverage the full potential of Large Language Models (LLMs) without advanced skills in crafting prompts. Our solution is specifically designed to bridge this divide, offering a tailored experience that seamlessly integrates with a student's life and academic demands.
RESEARCH & PLANNING
The simplest starting point was to survey our target users to learn about the deficiencies in current GPT and other AI models.
User Research - Questionnaire
Collaboratively design a questionnaire that analyzes pain points and potential needs/demands of UCSD CSE students in terms of their studying process (coursework learning challenges, academic knowledge acquisition, professional skills build-up, course and career planning, learning and networking resources, opportunity posts and feeds, etc. How can our app help to elevate their productivity and enhance their learning process?)
We chose to use Google Surveys to begin our user research.
How do you usually use these resources or apps for studying? What do you like/dislike about them? Which one of these resources or apps that you think most effectively assists with studying? And why? What do you think is missing from current student aid resources or apps?
User Research Survey (Link)
Clear trends began to emerge when we synthesized the data from our survey responses. Apps like ChatGPT were far and away the preferred means of assisting users with their customized needs. These are the biggest flaws that we identified:
-
While the UCSD academic plan webpage provides a basic framework for generating a four-year class schedule, its approach lacks the depth of customization that students truly need. A notable limitation is the generic representation of all elective courses simply as 'elective,' which misses the opportunity to guide students toward courses that align with their interests and career aspirations.
-
The need for specialized career and academic planning guidance. Students pursuing distinct paths, such as a professional career in software engineering or an academic research trajectory, require more than just a generic plan. They need a roadmap tailored to their specific ambitions, outlining what to study and when to study it throughout their college years.
-
Generate practice questions based on course content, or interactive reviews.
-
A more informative and navigable resource that connects students with academic professors/mentors and leaders in their fields of interest - Networking! (and schedule of classes being offered each quarter, prerequisites and a description of the course content).
CONFIDING: OUR SOLUTION
A pivotal consideration emerges in the design of the user interface (UI) to foster meaningful interaction. Unlike traditional content sharing, the challenge lies in balancing customization with the need to streamline the scenarios in which the LLM can be utilized effectively.
To navigate this, we propose a UI design that, while offering the customization students require, also incorporates a familiar dialogue-based interface reminiscent of OpenAI's question-answering format. This hybrid approach combines the intuitive, conversational engagement of a chat interface with enhanced features tailored to academic exploration.
In addition to traditional chatbot, the elements of our solution include:
Input for Customization
-
A large dataset containing previous course textbook, readings, slides, assignments, recordings.
-
Allow users to import academic materials or local information and data sets that need to be studied.
-
Provides file selection and drag-and-drop functionality.
-
Intelligent classification: Automatically identify the type and content of uploaded files and classify them into the correct course or learning module.
Feedback Mechanism
-
Direct Feedback Prompts: Periodically, or at the end of interactions, prompt users to rate their experience or provide feedback on the conversation.
-
Using thumbs up/down button at. the end of interaction, along with a pop up box to indicate explanations for our end.
Exam
Generations
-
Interactive "Quizlet" style: Users can input topics or select from predefined categories, the chatbot can present information in a flash card format. The chatbot then provides immediate feedback on correctness, along with explanations for further learning.
User Profile
-
User Profile feature: a streamlined option lets users input their grade, major, and career aspirations—a one-time setup for personalized interactions.
-
Task Manager: Help students plan learning tasks, course planning, etc.
-
Customized Career plan/Academic plan: A four-year roadmap tailored to students' goals.
-
A personalized plan outlining crucial courses, hands-on project opportunities, internships, and key certifications to pursue each year, aligning with industry standards
USER FLOW
As a team, we decided to test the idea of encouraging more interaction with the app by awarding points for certain actions, like file upload, personal profile, uploading course content and choose practice mode.
UI DESIGN PROCESS
HIGH FIDELITY PROTOTYPES
Chatbot
.png)
Login
