AI Literacy for Responsible Use
The project
Northstar is a conceptual EdTech app that helps users build practical AI skills through hands-on lessons tailored to their learning goals, designed for both regular AI users looking to use it more effectively and safely, and skeptics who aren't sure AI is worth the risk.
My role
UX/UI design, from research and strategy to interaction design, testing, and visual design.
Project duration
3 weeks
Tools
Figma
Perplexity AI (research)
Notion AI (transcription)
Why This Matters
The debate around AI adoption isn't slowing down, and neither is the rate at which people are jumping in without guidance. The more that people rely on AI without understanding how it works, the more they risk sharing sensitive information, producing outputs that they can't critically evaluate, or losing confidence in skills they used to trust.
The gap lies in AI education and how to use it well.
What Users Need
A product that meets them where they are, whether they use AI daily or avoid it entirely.
Practical skills for using AI effectively and safely, not just exposure to it.
A resource for staying informed as AI continues to evolve.
The solution
I designed an EdTech app that guides users through personalized, interactive lessons on how to use AI effectively, critically, and responsibly, with curated resources to stay informed.
Onboarding
Hands-on Lessons
AI Resources
Research
Competitive Analysis
My intake survey revealed that most people were learning through trial and error or simple Google searching, so I used that as a starting point to evaluate different learning methods.
The most notable finding was that structured online courses were the only method covering AI safety and regulation in any depth, but that content sat behind a paywall. Meanwhile, trial and error was the most common method people were actually using and the least equipped to teach responsible or critical use.
Learning from AI users and critics
With that framing in mind, I moved into user interviews to understand everyday AI use. I spoke with five participants ranging from daily AI users to active skeptics, covering their current habits, pain points, perceived benefits, and hopes and concerns about the future of AI.
Interview insights
Theme 1: AI is primarily used for productivity tasks like writing and summarizing, which focused my MVP on lessons built around those use cases.
Theme 2: Data privacy was the dominant concern across participants, which informed the decision to include a dedicated resource feature.
Theme 3: Some participants, particularly skeptics, worried that AI reliance was eroding their ability to think independently.
Theme 4: Regular AI users reported far fewer fears than expected, revealing that the design needed to serve meaningfully different mindsets.
My interviews surfaced two distinct relationships with AI that pulled in different directions. Rather than designing for an average user, I defined two archetypes to represent that tension and pressure-test my ideas against both throughout the process.
Defining the Problem
Across my research, a few threads kept surfacing: people were learning about AI informally and inconsistently, privacy concerns were widespread but poorly addressed, and skeptics and regular users alike were navigating AI without much support or structure.
How might we…
help users build practical AI skills while encouraging critical thinking and responsible use, empowering them to feel more in control of AI?
Design Strategy
Identifying core features
With the problem defined, I identified feature patterns across existing ed-tech platforms to understand what a learning experience typically requires. From there I made deliberate decisions about what to prioritize for an MVP, landing on three core areas:
Guided lessons
Teach foundational concepts through hands-on interaction.
Curated resources
Address the widespread need to stay informed about AI safety and developments.
Onboarding
Establish a personalized learning path from the start and frame the experience for both user types.
Mapping the experience
Sitemapping helped me draw clearer boundaries between the three core areas of the product, lessons, resources, and profile, which set a foundation for the designs. Mapping out user flows added another layer of detail, pushing me to think beyond the happy path and account for how users might move through the product in different ways.
Exploration
Low-fidelity wireframes
I developed low-fidelity wireframes across three core flows: onboarding, lessons, and navigation. The goal at this stage was to validate the overall structure and get a sense of how users perceived the product's value before investing in visual design.
Usability testing
Testing revealed that, even for a learning app, users expected the experience to be quick, digestible, and directive. The tension between an information-dense subject and a medium that demands brevity became a guiding constraint for future iterations.
Priority iterations
Iteration 1: Participants expected to interact with actual AI tools, not just answer multiple choice questions. Adding a hands-on AI experience gave the product more perceived value.
Iteration 2: Text density across screens needed to come down significantly to match users' expectations for a mobile learning experience.
Iteration 3: The resources section read as an academic reference page. Users expected something closer to a news feed with current, scannable content.
Iteration 4: Progress tracking was too vague. Users needed more context about their learning path and what was coming next.
Brand Design
Styling
I started with a moodboard of learning and AI product designs to understand the visual language of the category. Purple hues and sans serif typography came up consistently, pointing to a shared expectation around what an AI product should look and feel like.
After exploring several typefaces I landed on Instrument Sans, which balanced readability with a modern, approachable feel suited to a mobile learning experience.
Brand logo
The logo combines an open book with a north star, two symbols that reflect the product's core purpose of education and guidance. I wanted the mark to feel immediately legible as an education product while staying visually distinct enough to carry its own identity.
Refining the Experience
High-fidelity wireframes
Translating lo-fi wireframes into high-fidelity designs meant applying the brand system while navigating a specific tension: I wanted the interface to feel minimal and clean without feeling plain or visually unengaging. Most of my decisions at this stage were about finding that balance.
Hi-fi testing
I conducted a second round of testing with both new and returning participants to validate my lo-fi iterations and evaluate the overall cohesion of the design.
Key findings
Theme 1: Users successfully navigated core learning flows, confirming the app's structure matched common learning app mental models.
Theme 2: Testing revealed opportunities to reduce cognitive load through stronger visual hierarchy and lighter content density.
Theme 3: Participants responded positively to interactive learning elements, reinforcing the value of hands-on practice over passive reading.
Iterations
Reducing text
Continuing from lo-fi testing, text density remained an issue across several screens. Further trimming copy text helped the experience feel more appropriate for a mobile learning context.
Strengthening visual language
The onboarding flow in particular felt text-heavy and visually flat. Introducing icons and visual cues helped break up content and added context without adding cognitive load.
Improving hierarchy
The dashboard layout didn't match users' mental models, making it harder to find and access lesson information. Restructuring the hierarchy made the most important content more immediately accessible.
Final Design
What started as a text-heavy, multiple choice driven experience evolved into something more interactive, visually guided, and personalized. The final product reflects what users actually needed: a structured but digestible way to build confidence and competence with AI.
Define your learning path
Onboarding establishes a personalized starting point, addressing the need to serve two meaningfully different user mindsets from the first interaction.
Experience a lesson firsthand
Interactive lessons replaced passive multiple choice questions after lo-fi testing revealed users expected to actually practice using AI, not just answer questions about it.
Build skills
Hands-on exercises reinforce learning through doing, encouraging critical engagement rather than over-reliance on AI outputs.
Track growth and stay oriented
A restructured dashboard gives users clear visibility into their progress and what comes next, informed by hi-fi testing feedback on hierarchy.
Stay informed
Curated resources address the widespread concern around data privacy and AI safety that surfaced consistently across interviews.
Reflection
One of the most useful things this project revealed about my design process is that I have a tendency to tell rather than show. My early designs leaned heavily on text to explain the app, and it took testing feedback to push me toward visual patterns that communicated the same ideas more effectively. That's a lesson I'm carrying forward.
If I could revisit this project with more time, I would invest more deeply in the visual language. The time constraint meant leaning on familiar patterns that didn't fully capture the identity I had in mind for Northstar. I would have liked to develop a more considered design system that better matched the product's purpose and personality.
If I were to take this further, I would explore expanding interactive simulations for deeper learning, personalizing lesson plans based on user behavior, and integrating real-time AI feedback into exercises.
