X
celerator
celerator
celerator
X
X
U
U
U
Designing for a Brighter Future in Cognitive Science at UCSD
Designing for a Brighter Future in Cognitive Science at UCSD
Designing for a Brighter Future in Cognitive Science at UCSD
Timeline:
Timeline:
Mentor:
Team:
Tools:
Program:
Role:
Visual Designer
Up-Grade '25 by Design Co
10 Weeks
Alan Tran
Notebook LM, Figma, FigJam, Google Suite
Rayleen Marquez, Mia Jin, Alice Ma, Shreya Hiremath
Program:
Mentor:
Team:
Timeline:
Tools:
Role:
Visual Designer
Up-Grade '25 by Design Co
10 Weeks
Alan Tran
Rayleen Marquez, Mia Jin, Alice Ma, Shreya Hiremath
Notebook LM, Figma, FigJam, Google Suite
PROJECT OVERVIEW
PROJECT OVERVIEW
PROJECT OVERVIEW
Role:
Visual Designer
Program:
Up-Grade '25 by Design Co
Mentor:
Alan Tran
Team:
Rayleen Marquez, Mia Jin, Alice Ma, Shreya Hiremath
Timeline:
10 Weeks
Tools:
Notebook LM, Figma, FigJam, Google Suite
1
1
Goal:
Goal:
Conceptualize an LLM-based website that guides new (switched/entering/transfers) UCSD CogSci students in navigating the department.
Conceptualize an LLM-based website that guides new (switched/entering/transfers) UCSD CogSci students in navigating the department.
2
2
Research:
Research:
Learn more about LLM tools, select an AI tool for training, and remotely interview CogSci students/alumni to identify key product goals/pain points.
Learn more about LLM tools, select an AI tool for training, and remotely interview CogSci students/alumni to identify key product goals/pain points.
3
3
LLM Training:
LLM Training:
Feed the LLM files/documents/links related to UCSD CogSci courses, programs, and information.
Feed the LLM files/documents/links related to UCSD CogSci courses, programs, and information.
4
4
Design:
Design:
Create wireframes & hi-fi prototypes that visualize the functions of the LLM-based website if it were to exist.
Create wireframes & hi-fi prototypes that visualize the functions of the LLM-based website if it were to exist.
THE PROBLEM
THE PROBLEM
THE PROBLEM
UCSD Cognitive Science Students Lack Clear, Coherent Guidance, Leaving Them to Navigate the Major through Fragmented Resources and Peer Word-of-Mouth.
UCSD Cognitive Science Students Lack Clear, Coherent Guidance, Leaving Them to Navigate the Major through Fragmented Resources and Peer Word-of-Mouth.
UCSD Cognitive Science undergraduates often struggle to navigate their major because guidance is scattered across fragmented resources and informal peer word-of-mouth. To find even basic information, students are forced to juggle dozens of tabs, while key details such as course syllabi or project requirements are either buried or inaccessible to them. On top of this, limited access to upperclassmen and alumni makes it harder to find mentorship or practical advice. What should be a straightforward college planning experience has instead become a chaotic, time-consuming, and overwhelming process. These ongoing issues in the UCSD CogSci department led us to tackle the following problem statement:
UCSD Cognitive Science undergraduates often struggle to navigate their major because guidance is scattered across fragmented resources and informal peer word-of-mouth. To find even basic information, students are forced to juggle dozens of tabs, while key details such as course syllabi or project requirements are either buried or inaccessible to them. On top of this, limited access to upperclassmen and alumni makes it harder to find mentorship or practical advice. What should be a straightforward college planning experience has instead become a chaotic, time-consuming, and overwhelming process. These ongoing issues in the UCSD CogSci department led us to tackle the following problem statement:
How might we create a centralized website integrated with a powerful LLM to help UCSD Cognitive Science students easily access department resources, plan their academic path, and connect with alumni for potential career opportunities?
How might we create a centralized website integrated with a powerful LLM to help UCSD Cognitive Science students easily access department resources, plan their academic path, and connect with alumni for potential career opportunities?
OUR SOLUTION
OUR SOLUTION
OUR SOLUTION
An All-in-One, AI-Assisted Website, Exclusively Designed for Cognitive Science Students at UCSD.
An All-in-One, AI-Assisted Website, Exclusively Designed for Cognitive Science Students at UCSD.
An All-in-One, AI-Assisted Website, Exclusively Designed for Cognitive Science Students at UCSD.
Our team's vision for this product pitch was to provide a seamless and informative experience for our incoming Cognitive Science majors at UCSD. To achieve this, we prototyped a centralized website concept integrated with Notebook LM that can assist its users from anything ranging from course planning to gaining real alumni advice.
Our team's vision for this product pitch was to provide a seamless and informative experience for our incoming Cognitive Science majors at UCSD. To achieve this, we prototyped a centralized website concept integrated with Notebook LM that can assist its users from anything ranging from course planning to gaining real alumni advice.
Our team's vision for this product pitch was to provide a seamless and informative experience for our incoming Cognitive Science majors at UCSD. To achieve this, we prototyped a centralized website concept integrated with Notebook LM that can assist its users from anything ranging from course planning to gaining real alumni advice.
USER RESEARCH FINDINGS
USER RESEARCH FINDINGS
USER RESEARCH FINDINGS
Understanding Our Users
Understanding Our Users
We conducted in-depth, online user interviews on 15 UCSD Cognitive Science majors to identify how they navigated the major and what resources they used to guide them. Then, we performed affinity mapping to organize our insights into common themes, which helped us highlight the interviewees' key pain points and necessities while they tried to find out more about the field.
We conducted in-depth, online user interviews on 15 UCSD Cognitive Science majors to identify how they navigated the major and what resources they used to guide them. Then, we performed affinity mapping to organize our insights into common themes, which helped us highlight the interviewees' key pain points and necessities while they tried to find out more about the field.
Participant Demographics
Participant Demographics
We aimed to target Cognitive Science undergraduates at UCSD who can use our product to assist them in planning classes, finding course syllabi, and gaining alumni/professional advice. Mainly, our focus was on newer students who switched majors, transferred, or are just starting their first year in Cognitive Science. Therefore, we attempted to interview a higher percentage of CogSci undergraduates than alumni in our user research.
We aimed to target Cognitive Science undergraduates at UCSD who can use our product to assist them in planning classes, finding course syllabi, and gaining alumni/professional advice. Mainly, our focus was on newer students who switched majors, transferred, or are just starting their first year in Cognitive Science. Therefore, we attempted to interview a higher percentage of CogSci undergraduates than alumni in our user research.
Most Helpful Resources
Most Helpful Resources
On average, the participants rated the current resources a 74% in terms of effectiveness.
On average, the participants rated the current resources a 74% in terms of effectiveness.
"It's difficult to find information and more people that are actually in that major. I think that's why it became a little more difficult because I could look up a bio major or something… and millions of things pop up but for Cognitive Science, it's a little more restricted."
"It's difficult to find information and more people that are actually in that major. I think that's why it became a little more difficult because I could look up a bio major or something… and millions of things pop up but for Cognitive Science, it's a little more restricted."
80% of participants experienced pain points with navigating UCSD Cognitive Science-related resources, the top issues being…
80% of participants experienced pain points with navigating UCSD Cognitive Science-related resources, the top issues being…
Insufficient information clarity & accessibility.
Challenges in course planning & scheduling.
Lack of practical relevance/hands-on experience in some courses.
Difficulty finding career opportunities and alumni connections.
Insufficient information clarity & accessibility.
Challenges in course planning & scheduling.
Lack of practical relevance/hands-on experience in some courses.
Difficulty finding career opportunities and alumni connections.
PRODUCT CONSIDERATIONS
PRODUCT CONSIDERATIONS
PRODUCT CONSIDERATIONS
Learning from the Past. Designing for the People.
Learning from the Past. Designing for the People.
Create a stress-free, helpful resource for new CogSci students at UCSD.
Everything should be accessible in one place. (No more screens filled with numerous tabs!)
Integrate a personalized course planner that recommends classes based on your interest and learning preferences.
Provide an FAQ feature that also allows users to engage with alumni and ask them industry-related questions.
Create a stress-free, helpful resource for new CogSci students at UCSD.
Everything should be accessible in one place. (No more screens filled with numerous tabs!)
Integrate a personalized course planner that recommends classes based on your interest and learning preferences.
Provide an FAQ feature that also allows users to engage with alumni and ask them industry-related questions.
LLM RESEARCH
LLM RESEARCH
LLM RESEARCH
Finding the Best AI Tool
Finding the Best AI Tool
I ran a competitive analysis on three AI tools (Notebook LM, Claude, and Perplexity) to find the right one to implement into our prospective website. Comparing their qualities allowed my team and I to assess each LLM's strengths and weaknesses, ultimately helping us decide which tool is best-suited to combat our problem space. While performing our research, we wanted to ensure the LLM we selected would meet the following criteria:
I ran a competitive analysis on three AI tools (Notebook LM, Claude, and Perplexity) to find the right one to implement into our prospective website. Comparing their qualities allowed my team and I to assess each LLM's strengths and weaknesses, ultimately helping us decide which tool is best-suited to combat our problem space. While performing our research, we wanted to ensure the LLM we selected would meet the following criteria:
I ran a competitive analysis on three AI tools (Notebook LM, Claude, and Perplexity) to find the right one to implement into our prospective website. Comparing their qualities allowed my team and I to assess each LLM's strengths and weaknesses, ultimately helping us decide which tool is best-suited to combat our problem space. While performing our research, we wanted to ensure the LLM we selected would meet the following criteria:
Does the LLM provide accurate and helpful student advice?
Does it have low hallucination rates to reduce misinformation?
Is the tool collaborative and shareable across users?
Does the LLM provide accurate and helpful student advice?
Does it have low hallucination rates to reduce misinformation?
Is the tool collaborative and shareable across users?
And The Winner is….
And The Winner is….
Notebook LM!
Notebook LM!
But Why?
But Why?
We finalized our decision on Notebook LM because it met all three of our criteria from above. Its clean, notebook-style format makes the LLM convenient for students to use, and its collaborative features are perfect for sharing with peers. Unlike other AI chatbots, Notebook LM has one of the lowest hallucination rates since it only generates responses based on the sources it is fed, making it less susceptible in providing non-existent or false information to students. Overall, its accuracy, simplicity and team-friendly qualities made it the best candidate for our product.
We finalized our decision on Notebook LM because it met all three of our criteria from above. Its clean, notebook-style format makes the LLM convenient for students to use, and its collaborative features are perfect for sharing with peers. Unlike other AI chatbots, Notebook LM has one of the lowest hallucination rates since it only generates responses based on the sources it is fed, making it less susceptible in providing non-existent or false information to students. Overall, its accuracy, simplicity and team-friendly qualities made it the best candidate for our product.
Now, it's Time to Train.
Now, it's Time to Train.
1
1
Started curating any accessible course syllabi/project guides for training.
Started curating any accessible course syllabi/project guides for training.
2
2
Compiled CogSci course information into grouped documents to save training capacity.
Compiled CogSci course information into grouped documents to save training capacity.
3
3
Inserted relevant links into Notebook LM from the official UCSD website and CogSci Department.
Inserted relevant links into Notebook LM from the official UCSD website and CogSci Department.
A Glimpse of What Notebook LM Can Do…
A Glimpse of What Notebook LM Can Do…
A Glimpse of What Notebook LM Can Do…
Get Help Selecting a CogSci Specialization That Suits Your Interests!
Get Help Selecting a CogSci Specialization That Suits Your Interests!
Based on your preferences, Notebook LM can guide you in choosing the Cognitive Science specialization that most aligns with your current interests and even future ambitions.
Based on your preferences, Notebook LM can guide you in choosing the Cognitive Science specialization that most aligns with your current interests and even future ambitions.
Conflicted Between Which Course to Choose? No Problem!
Conflicted Between Which Course to Choose? No Problem!
With its insider knowledge, Notebook LM can provide in-depth explanations on the classes of your choosing to help you find the perfect fit for you.
With its insider knowledge, Notebook LM can provide in-depth explanations on the classes of your choosing to help you find the perfect fit for you.
Don't Know the Difference Between Getting a B.A. or B.S. in CogSci? We Got You!
Don't Know the Difference Between Getting a B.A. or B.S. in CogSci? We Got You!
Notebook LM can clear your confusion between the two CogSci degrees in a matter of seconds. It breaks down each degree based on its graduation requirements and differentiating features to assist you in your selection.
Notebook LM can clear your confusion between the two CogSci degrees in a matter of seconds. It breaks down each degree based on its graduation requirements and differentiating features to assist you in your selection.
DESIGN EXECUTION
DESIGN EXECUTION
DESIGN EXECUTION
(Due to the research-intensive nature of this project and the short program timeline, our team moved directly into designing the mid-fidelity wireframes of the website in Figma. While this approach minimized more thoughtful time spent on early ideation, it enabled us to align our designs more efficiently with the insights gathered during our research.)
(Due to the research-intensive nature of this project and the short program timeline, our team moved directly into designing the mid-fidelity wireframes of the website in Figma. While this approach minimized more thoughtful time spent on early ideation, it enabled us to align our designs more efficiently with the insights gathered during our research.)
Mid-Fidelity Wireframes
Mid-Fidelity Wireframes
While brainstorming through the design process, we decided to split the website contents into four parts based on the top pain points expressed by our interviewees: advising assistance, alumni advice, four-year planning, and course information. Our team envisioned setting our AI-integrated website apart from its other competitors, hoping to expand beyond the typical chat-bot feature. Hence, we designed each section to be explored by its users in slightly different ways:
While brainstorming through the design process, we decided to split the website contents into four parts based on the top pain points expressed by our interviewees: advising assistance, alumni advice, four-year planning, and course information. Our team envisioned setting our AI-integrated website apart from its other competitors, hoping to expand beyond the typical chat-bot feature. Hence, we designed each section to be explored by its users in slightly different ways:
Gaining Advice
Gaining Advice
The advising assistant would follow the usual chat-bot format to keep the interaction between the AI and user conversational.
The advising assistant would follow the usual chat-bot format to keep the interaction between the AI and user conversational.
Quarter Scheduling
Quarter Scheduling
Alternately, the four-year plan tab would be formatted as a form to keep the planning process concise and straightforward. Three parts and you have a custom outline!
Alternately, the four-year plan tab would be formatted as a form to keep the planning process concise and straightforward. Three parts and you have a custom outline!
Connecting with Alumni
Connecting with Alumni
The alumni advice tab would also be a chat-bot feature, mixed with a common FAQ section that users can utilize as quick access.
The alumni advice tab would also be a chat-bot feature, mixed with a common FAQ section that users can utilize as quick access.
Course Familiarization
Course Familiarization
Finally, the course information tab would follow a categorical layout, including customization buttons for selection to make the process less overwhelming for users.
Finally, the course information tab would follow a categorical layout, including customization buttons for selection to make the process less overwhelming for users.
THE RESULTS
THE RESULTS
THE RESULTS
(The following website demonstrations are prototyped concepts, and a fully-developed site does not exist.)
(The following website demonstrations are prototyped concepts, and a fully-developed site does not exist.)
Stuck on a Project, Paper, or Course Planning? Talk to Our AI-Advising Assistant!
Stuck on a Project, Paper, or Course Planning? Talk to Our AI-Advising Assistant!
Wish to Receive a Little Extra Wisdom? Gain Alumni Advice Through FAQs and Your Own Personalized Questions.
Wish to Receive a Little Extra Wisdom? Gain Alumni Advice Through FAQs and Your Own Personalized Questions.
Find the Right Four-Year Plan For You By Selecting Your Specialization, UCSD College, and Class Preferences.
Find the Right Four-Year Plan For You By Selecting Your Specialization, UCSD College, and Class Preferences.
Quickly Access Course Information Based on Your Specialization!
Quickly Access Course Information Based on Your Specialization!
CONSTRAINTS & JUSTIFICATION
CONSTRAINTS & JUSTIFICATION
CONSTRAINTS & JUSTIFICATION
LLM Limitations
LLM Limitations
Notebook LM Does Not Have a Public API Key.
1.
Notebook LM does not have a public API key.
This restriction prevented us from developing a fully functional external website integrated with Notebook LM, which is why we focused on creating a conceptual prototype instead.
This restriction prevented us from developing a fully functional external website integrated with Notebook LM, which is why we focused on creating a conceptual prototype instead.
We Trained the LLM Using its Free Version.
2.
We trained the LLM using its free version.
The free version limits both the number of sources (up to 50) and the size of each source (max 500,000 words), which required us to compile and organize content into their respective documents to optimize the provided space.
The free version limits both the number of sources (up to 50) and the size of each source (max 500,000 words), which required us to compile and organize content into their respective documents to optimize the provided space.
Team Limitations
Team Limitations
LLM Training Sources are Limited to What the Team Members Can Access.
3.
LLM Training Sources are Limited to What the Team Members Can Access.
Certain courses have more available resources than others, so the LLM may not cover every Cognitive Science course comprehensively, causing some classes to have more saturated information.
Certain courses have more available resources than others, so the LLM may not cover every Cognitive Science course comprehensively, causing some classes to have more saturated information.
Our User Research is Not Fully Representative of the Target Audience.
4.
Our User Research is Not Fully Representative of the Target Audience.
Our online interviews included a small sample that heavily represented the design and interaction specialization, which may have introduced some bias into our research outcomes.
Our online interviews included a small sample that heavily represented the design and interaction specialization, which may have introduced some bias into our research outcomes.
Program Limitations
Program Limitations
We are Limited to a 10-Week Timeline.
5.
We are Limited to a 10-Week Timeline.
The research-heavy nature of the project and the novelty of the AI tool meant progress was constrained by time, but we were still able to make meaningful advances within this period.
The research-heavy nature of the project and the novelty of the AI tool meant progress was constrained by time, but we were still able to make meaningful advances within this period.
We Cannot Invest Money into the Project.
6.
We Cannot Invest Money into the Project.
Lack of financial resources limited access to additional data, training materials, and tools that could have accelerated development. Additionally, since the project is not ongoing, there is no guarantee that a future team would choose to invest funds toward its completion.
Lack of financial resources limited access to additional data, training materials, and tools that could have accelerated development. Additionally, since the project is not ongoing, there is no guarantee that a future team would choose to invest funds toward its completion.
REFLECTION
REFLECTION
REFLECTION
1
Expand Beyond Your Comfort Zone.
Expand Beyond Your Comfort Zone.
Upon initial understanding of the product pitch introduced to our team by our mentor, we were intimidated by the prospect of working with something as novel as artificial intelligence. Despite our uncertainty, we were up for the challenge because we unanimously agreed that the purpose of us participating in this program was to grow our knowledge and skills, not just continue working with tasks we have done in the past. Now, I not only familiarized myself with the once alien term, LLM, but know how to train one!
Upon initial understanding of the product pitch introduced to our team by our mentor, we were intimidated by the prospect of working with something as novel as artificial intelligence. Despite our uncertainty, we were up for the challenge because we unanimously agreed that the purpose of us participating in this program was to grow our knowledge and skills, not just continue working with tasks we have done in the past. Now, I not only familiarized myself with the once alien term, LLM, but know how to train one!
2
You Can Never Do Too Much Research.
You Can Never Do Too Much Research.
This project was different from ones I've worked on in the past due to its heavy emphasis on research. Unlike the usual 2-3 weeks of research we would perform, this pitch required us to dedicate as much as 7 weeks out of our 10-week timeline. Yet, we soon came to realize that spending more time thoroughly understanding our project and the tools we are working with made the rest of the design process go more smoothly. Since working with an LLM was new to all of us, extended research was a must for us to create an enticing product.
This project was different from ones I've worked on in the past due to its heavy emphasis on research. Unlike the usual 2-3 weeks of research we would perform, this pitch required us to dedicate as much as 7 weeks out of our 10-week timeline. Yet, we soon came to realize that spending more time thoroughly understanding our project and the tools we are working with made the rest of the design process go more smoothly. Since working with an LLM was new to all of us, extended research was a must for us to create an enticing product.
3
Sometimes, Progress is Enough.
Sometimes, Progress is Enough.
As mentioned before, our team was only able to design a visual conceptualization of UXcelerator. However, our incomplete project does not mean we did not accomplish anything. Despite our setbacks, we were able to work on the beginning of something new, a project that can be taken on by future designers and advanced to even better improvements than we imagined. We hope that the progress we made (no matter how little) will spark continual development of our ideas, eventually forming a revolutionizing product.
As mentioned before, our team was only able to design a visual conceptualization of UXcelerator. However, our incomplete project does not mean we did not accomplish anything. Despite our setbacks, we were able to work on the beginning of something new, a project that can be taken on by future designers and advanced to even better improvements than we imagined. We hope that the progress we made (no matter how little) will spark continual development of our ideas, eventually forming a revolutionizing product.
Potential Next Steps
Potential Next Steps
Potential Next Steps
Develop an existing, public website for UXcelerator: Moving forward, we would have liked to build a fully functional website that integrates Notebook LM, allowing students to access all relevant resources, course information, and guidance in one centralized platform. This would transform our conceptual prototype into a tangible product that could directly support UCSD Cognitive Science undergraduates.
Develop an existing, public website for UXcelerator: Moving forward, we would have liked to build a fully functional website that integrates Notebook LM, allowing students to access all relevant resources, course information, and guidance in one centralized platform. This would transform our conceptual prototype into a tangible product that could directly support UCSD Cognitive Science undergraduates.
Conduct user testing of UXcelerator: To test the intuitiveness and effectiveness of the platform, it would be important to run trials of it with a diverse group of CogSci students. Gathering feedback on usability, navigation, and content clarity would help refine the interface and improve the overall user experience, ensuring we are designing for the people.
Conduct user testing of UXcelerator: To test the intuitiveness and effectiveness of the platform, it would be important to run trials of it with a diverse group of CogSci students. Gathering feedback on usability, navigation, and content clarity would help refine the interface and improve the overall user experience, ensuring we are designing for the people.
Create a marketing and outreach strategy for UXcelerator to reach UCSD CogSci students: Raising awareness of the platform is crucial for its implementation. Developing a targeted strategy through student groups, UCSD CogSci-affiliated organizations, and social media would help ensure that the tool reaches its intended audience and achieves meaningful engagement.
Create a marketing and outreach strategy for UXcelerator to reach UCSD CogSci students: Raising awareness of the platform is crucial for its implementation. Developing a targeted strategy through student groups, UCSD CogSci-affiliated organizations, and social media would help ensure that the tool reaches its intended audience and achieves meaningful engagement.
