A passion project in which I explore Duolingo, its strengths and limitations, and offer a feature addition to help users strengthen their conversational speaking skills. Language learners can practice listening, thinking, and speaking in their target language.
View DesignStephanie Lin (Product Designer)
80hrs / 1month
Imagine shifting the focus of Duolingo from memorizing vocabulary to encouraging thought formation and speaking for language learners.
Please Note: During the development of this project, Duolingo released a similar feature within their paid tier, DuolingoMax known as "DuoRoleplaying" and "Explain my Answer".
Language learners felt limited by features available and weren’t confident in their conversational abilities.
An AI "Teacher" (Duo) that grades the user's speaking or writing, noting suggestions for improvements.
Figure 1. Highlights of DuoNarratives Feature: Overview, Feedback Summary, and Detailed Feedback
Figure 2. Design process for DuoNarratives
Rather than additional community support, users felt that Duolingo was ineffective in facilitating their ability to engage in conversation in a new language.
Learning struggles in the community were complex, with diverse trends observed among heritage and non-heritage learners based on proficiency.
Prompts had heavy emphasis on memorization and recognition. Users felt there was more emphasis on gamification rather than learning.
Despite their Duolingo engagement and progress, users faced challenges with real-time understanding of spoken words and generating fluent responses.
Image 1. Insights from a 3-Month Long Duolingo User: Importance of Language Engagement in Skill Building
After getting a better understanding of the problem, I became curious. I spoke with a speech language pathologist to understand: How does learning a language work?
Figure 3. Progression of Speaking a New Language
I newly hypothesized that people struggled to speak in a language because they weren't engaged in activities that helped them "think" in the target language.
I confirmed my hypothesis and learned that its also important to provide progression when it comes to language learning. It's not easy to go from a simple word to a full sentence, there needs to be an intermediary step(see Figure 1). When I referenced my interview, all participants struggled with the muscle of thinking of simple sentences in a new language. Duolingo is currently set up for recognition.
I also gathered and analyzed Duolingo’s branding and company values to help me understand what makes Duolingo memorable and successful.
Figure 4. Intersection of UX and Brand Design: Duolingo’s Use of Branding to Create an Immersive Learning Experience
Duolingo's screens feature vibrant colors, simple characters, rounded edges, concise copy, and a clear hierarchy, allowing for easy comprehension at a glance, while also ensuring that most screens are accessible without scrolling. In terms of features, there seems to be a trend of sunsetting community-features including forums and duels and releasing features that support an individual learner’s journey such as the new homepage with its directed lesson map.
One important trend that guided future decision making was that users were not confident when they practicing aloud.
Figure 5. Brief Exploration of Duolingo’s Diverse Learner Archetypes with Solo and Collaborative Learners
While users had differences when it came to their style of learning: collaborative versus individual environments. Users shared similar goals and pain points.
Users feel insecure and uncertain about practicing a new language aloud.
Feel confident in having a conversation in a new language
With my workshop participants, we explored “How might we create effective study spaces within Duolingo?”
Image 2. Exploring Potential Feature Additions to Duolingo in an Ideation Workshop
I hosted a 1-hour workshop with a few UX designers and developers. First, I debriefed my participants. Then, we did a quick 5-minute brainstorming activity followed up a 25-minute in-depth activity where participants explored an idea they thought was promising. This in-depth activity (pictured above) included 1) the big idea, 2) important details, 3) hypothesis/impact, and 4) potential challenges. I ended with a wrap-up discussion where we discussed feasibility and effort.
Based on my research, I chose "DuoNarratives" because it best fit existing Duolingo business and design efforts. The following key traits of DuoNarratives also mirror existing efforts to improve their service.
Added personal prompts to create an immersive and thought-provoking learning experience
Integrated a system of increasing difficulty to help users build the skills necessary to have a conversation
Promote self-awareness and learning through error corrections can help learners improve their weaknesses
I used wireflows to identify components and patterns I could use from the existing Duolingo system and which components and patterns I needed to create. After I finalized the wireflows, my previous research on Duolingo's design system allowed me to seamlessly move towards hi-fidelity and prototyping.
Image 3. Sketching Wireflows for DuoNarratives
Testing revealed many weaknesses in the design, with the biggest issue being that users (n=5) could not accurately interpret the feedback from AI. This issue was the most complex and required me to explore potential solutions in low-fidelity wireframes.
of users were able to complete the feedback task in the first pass.
of users completed Task 2 after receiving a debrief.
of users believed the feature concept would be effective for learning.
Image 4. Insights from A DuoNarrative Test Participant
I conducted moderated usability tests with 5 participants over Zoom. Each session lasted approximately 45-minutes. Participants were asked to located the new feature and accurately interpret the feedback from AI. Users were tested in Mandarin (Task 1) and Spanish (Task 2). They were not required to have knowledge of either as the goal was to test the design and not the nature of the content, though language content errors did add some levels of confusion.
Figure 6. Confusing Speech Components Discovered During Usability Tests
Figure 7. Confusing Speech Components Discovered During Usability Tests
The most critical usability issue was the heavy cognitive load on users. This problem was complex since there were multiple parts adding to their cognitive load (see Issue 1 and 2 above) as well as the overwhelming amount of information provided on the feedback screen, language content errors, and the constraints of the existing design system. I used lo-fi to help me explore alternatives quickly.
Image 5. Simplifying color codes within Detail Feedback Screen
Image 6. Simplifying color codes and adding an accordion within Detail Feedback Screen
Image 7. Adding a feedback overview and brief feedback screen before Detail Feedback Screen
