Jan 2025
Jan 2025
Converse in Japanese
Converse in Japanese
Company: iago
Project Type: Iteration on AI chat interface to improve usability
Role: Sole product designer
iago is a mobile app that teaches Japanese through immersion by having learners navigate a visual novel game and speaking from day one. In each situation, learners participate in AI prompt-guided chats that increase in difficulty, eventually forming a full conversation at the end.
iago is a mobile app that teaches Japanese through immersion by having learners navigate a visual novel game and speaking from day one. In each situation, learners participate in AI prompt-guided chats that increase in difficulty, eventually forming a full conversation at the end.
iago is a mobile app that teaches Japanese through immersion by having learners navigate a visual novel game and speaking from day one. In each situation, learners participate in AI prompt-guided chats that increase in difficulty, eventually forming a full conversation at the end.
Problem
Learners aren't motivated to complete Chat exercises, despite their value in reinforcing concepts and building skills to converse in Japanese.
Learners aren't motivated to complete Chat exercises, despite their value in reinforcing concepts and building skills to converse in Japanese.
Approach
Improve the UX experience by adding display modes options, enabling message edits with auto-submitted feedback, and moving hints inline with the conversation.
Improve the UX experience by adding display modes options, enabling message edits with auto-submitted feedback, and moving hints inline with the conversation.
Results
Increased chat usage by 24% and daily study time by 20%, improving learners' potential to better converse in Japanese and dedication to studying.
01
Discovery
01
Discovery
01
Discovery
About four months prior, the iago curriculum transitioned from an open-world format to a more structured visual novel experience. During this pivot, the Stories (lessons) and Practices were redesigned to align with the new structure. However, the Chats feature remained unchanged and more than 43% of learners were skipping Chats altogether .
To support a more holistic learning journey, the Iago team identified the need to better align the app with Kolb’s Learning Cycle. The Chat experience is crucial for providing Concrete Experiences—realistic, interactive moments that simulate life-like dialogue. Enhancing the Chat feature became the next priority to ensure that the entire learning cycle is represented within the app.
About four months prior, the iago curriculum transitioned from an open-world format to a more structured visual novel experience. During this pivot, the Stories (lessons) and Practices were redesigned to align with the new structure. However, the Chats feature remained unchanged and more than 43% of learners were skipping Chats altogether .
To support a more holistic learning journey, the Iago team identified the need to better align the app with Kolb’s Learning Cycle. The Chat experience is crucial for providing Concrete Experiences—realistic, interactive moments that simulate life-like dialogue. Enhancing the Chat feature became the next priority to ensure that the entire learning cycle is represented within the app.
Kolb's Learning Cycle
Kolb's Learning Cycle
This was summarized into the following problem statement:
“
Learners aren't motivated to complete Chat exercises, despite their value in reinforcing concepts and building skills to converse in Japanese.
By improving and elevating the Chat experience, we may be able to encourage greater learner engagement—which could meaningfully enhance both knowledge retention and long-term commitment to the learning journey.
This was summarized into the following problem statement:
“
Learners aren't motivated to complete Chat exercises, despite their value in reinforcing concepts and building skills to converse in Japanese.
By improving and elevating the Chat experience, we may be able to encourage greater learner engagement—which could meaningfully enhance both knowledge retention and long-term commitment to the learning journey.
02
Research
02
Research
02
Research
To inform the redesign, I conducted a heuristic analysis of both direct ed-tech competitors and relevant apps in the productivity and communication space. I focused on voice recording, chat interfaces, and audio transcription patterns to identify effective design solutions.
These patterns highlighted the need for more robust controls within our chat experience—especially around managing voice input—and clearer guidance for users on what to say.
To inform the redesign, I conducted a heuristic analysis of both direct ed-tech competitors and relevant apps in the productivity and communication space. I focused on voice recording, chat interfaces, and audio transcription patterns to identify effective design solutions.
These patterns highlighted the need for more robust controls within our chat experience—especially around managing voice input—and clearer guidance for users on what to say.
Speak
Speak
In-line hinting
In-line hinting
Facebook Messenger
Facebook Messenger
Edit messages
Edit messages
Google Gemini
Google Gemini
Live audio transcript
Live audio transcript
Next was internal research, which was mainly conducted by sifting through the bug reports and collecting comments on Iago’s social media channels. The majority of comments touched upon these three issues:
Next was internal research, which was mainly conducted by sifting through the bug reports and collecting comments on Iago’s social media channels. The majority of comments touched upon these three issues:
03
Ideate
03
Ideate
03
Ideate
After completing competitive research, I facilitated a small brainstorming session with our frontend engineers and community manager. We used previously developed user stories to ensure that all ideas directly addressed the needs of our specific users.
From this session, a range of ideas emerged, which I grouped into three key themes. To address gaps in the current Chat experience holistically, I selected one representative solution from each theme to move forward with.
After completing competitive research, I facilitated a small brainstorming session with our frontend engineers and community manager. We used previously developed user stories to ensure that all ideas directly addressed the needs of our specific users.
From this session, a range of ideas emerged, which I grouped into three key themes. To address gaps in the current Chat experience holistically, I selected one representative solution from each theme to move forward with.
01
Edit Response
Changed recording behavior from hold to record to tap to start/stop recording, confirm message to send.
WHY: Learners can confirm that their responses were accurately recorded or attempt again. Unintended responses will not be sent and run the chat off-course by the AI.
01
Edit Response
Changed recording behavior from hold to record to tap to start/stop recording, confirm message to send.
WHY: Learners can confirm that their responses were accurately recorded or attempt again. Unintended responses will not be sent and run the chat off-course by the AI.
01
Edit Response
Changed recording behavior from hold to record to tap to start/stop recording, confirm message to send.
WHY: Learners can confirm that their responses were accurately recorded or attempt again. Unintended responses will not be sent and run the chat off-course by the AI.
02
Switch Display Mode
Added ability to change display mode (Japanese Writing System) for specific messages.
WHY: Allows the user to challenge themselves with Kanji + Kana if they want or easily revert back.
02
Switch Display Mode
Added ability to change display mode (Japanese Writing System) for specific messages.
WHY: Allows the user to challenge themselves with Kanji + Kana if they want or easily revert back.
02
Switch Display Mode
Added ability to change display mode (Japanese Writing System) for specific messages.
WHY: Allows the user to challenge themselves with Kanji + Kana if they want or easily revert back.
03
In-Line Hinting
Added ability to change display mode (Japanese Writing System) for each chat message.
WHY: Allows the user to challenge themselves with Kanji + Kana if they want or easily revert back.
03
In-Line Hinting
Added ability to change display mode (Japanese Writing System) for each chat message.
WHY: Allows the user to challenge themselves with Kanji + Kana if they want or easily revert back.
03
In-Line Hinting
Added ability to change display mode (Japanese Writing System) for each chat message.
WHY: Allows the user to challenge themselves with Kanji + Kana if they want or easily revert back.
04
Reiterate
04
Reiterate
04
Reiterate
After a discussion with the co-CEO/frontend engineer, he decided to reduce the project size to two weeks due to limited engineering resources. By bumping up the timeline, we had to ruthlessly prioritize solutions that required the least amount of changes that resulted in short term wins.
After a discussion with the co-CEO/frontend engineer, he decided to reduce the project size to two weeks due to limited engineering resources. By bumping up the timeline, we had to ruthlessly prioritize solutions that required the least amount of changes that resulted in short term wins.
01 V2
Edit Typed Response
Allow users to manually override their recorded responses by submitting typed responses. A ticket is also automatically submitted.
WHY: Circumvent voice recognition limitations while still giving users the opportunity to learn.
Before
After
01 V2
Edit Typed Response
Allow users to manually override their recorded responses by submitting typed responses. A ticket is also automatically submitted.
WHY: Circumvent voice recognition limitations while still giving users the opportunity to learn.
Before
After
02 V2
Change Display Mode
Change overall display mode of chat instead of individual messages.
WHY: Reuse designs from account settings and reduce engineering lift.
Before
After
02 V2
Change Display Mode
Change overall display mode of chat instead of individual messages.
WHY: Reuse designs from account settings and reduce engineering lift.
Before
After
02 V2
Change Display Mode
Change overall display mode of chat instead of individual messages.
WHY: Reuse designs from account settings and reduce engineering lift.
Before
After
01 V2
Edit Typed Response
Allow users to manually override their recorded responses by submitting typed responses. A ticket is also automatically submitted.
WHY: Circumvent voice recognition limitations while still giving users the opportunity to learn.
Before
After
01 V2
Edit Typed Response
Allow users to manually override their recorded responses by submitting typed responses. A ticket is also automatically submitted.
WHY: Circumvent voice recognition limitations while still giving users the opportunity to learn.
Before
After
03 V2
Manual In-line Hints
Content team manually sifted through difficult prompts to adjust wording and add true hints.
WHY: Avoid huge changes to entire page layout. Stopgap solution due to AI limitations to give true hints.
Before
After
03 V2
Manual In-line Hints
Content team manually sifted through difficult prompts to adjust wording and add true hints.
WHY: Avoid huge changes to entire page layout. Stopgap solution due to AI limitations to give true hints.
Before
After
03 V2
Manual In-line Hints
Content team manually sifted through difficult prompts to adjust wording and add true hints.
WHY: Avoid huge changes to entire page layout. Stopgap solution due to AI limitations to give true hints.
Before
After
05
Results
05
Results
05
Results
After two weeks, we observed a 24% increase in overall Chat usage. Average session length also rose—from approximately 15 minutes to 18 minutes—while the average number of days to complete a situation remained steady at 5 days for Chat users. In other words, learners are now spending more time actively engaged in their studies.
After two weeks, we observed a 24% increase in overall Chat usage. Average session length also rose—from approximately 15 minutes to 18 minutes—while the average number of days to complete a situation remained steady at 5 days for Chat users. In other words, learners are now spending more time actively engaged in their studies.
An increase in Chat use, an increase in overall studying
An increase in Chat use, an increase in overall studying
+24%
+24%
chat use
chat use
+25%
+25%
daily study time
daily study time
While the new Chat interface has only been live for two weeks (too early to determine its long-term impact on learner retention), we have already received enthusiastic feedback from our learners. Many have expressed appreciation for the improvements in aiding their speaking practice, and we anticipate deeper, more sustained Japanese learning as learners continue to engage in more conversations!
While the new Chat interface has only been live for two weeks (too early to determine its long-term impact on learner retention), we have already received enthusiastic feedback from our learners. Many have expressed appreciation for the improvements in aiding their speaking practice, and we anticipate deeper, more sustained Japanese learning as learners continue to engage in more conversations!
“
I love how iago is more focused on understanding and applying; the chat really recognized my mistakes so I know how I can improve on my speaking."
Kara K.
“
I love how iago is more focused on understanding and applying; the chat really recognized my mistakes so I know how I can improve on my speaking."
Kara K.
“
I love how iago is more focused on understanding and applying; the chat really recognized my mistakes so I know how I can improve on my speaking."
Kara K.
Takeaways
Takeaways
01
Loop engineers in sooner
Loop engineers in sooner
AI voice recognition limitations required front-end workarounds, but late discussions led to infeasible designs. Had I shared mid-fidelity designs earlier with engineering, these constraints could have been surfaced sooner.
AI voice recognition limitations required front-end workarounds, but late discussions led to infeasible designs. Had I shared mid-fidelity designs earlier with engineering, these constraints could have been surfaced sooner.
02
More guidance != better
More guidance != better
As designers, we aim to prevent user mistakes, but in learning, too much leniency can backfire. The AI chat’s flexibility frustrated users, preventing them from improving. I lost sight of our users’ needs.
As designers, we aim to prevent user mistakes, but in learning, too much leniency can backfire. The AI chat’s flexibility frustrated users, preventing them from improving. I lost sight of our users’ needs.
Next Steps…
Next Steps…
Monitor user reports
Monitor user reports
Track chat report trends. Conduct usability testing if needed.
Track chat report trends. Conduct usability testing if needed.
Rework Hinting
Rework Hinting
Redesign in-line hinting to accommodate AI improvements for personalized hints.
Redesign in-line hinting to accommodate AI improvements for personalized hints.
Summarization Screen
Summarization Screen
Design a post-Chat summary page to prompt user reflection after completing Conversations.
Design a post-Chat summary page to prompt user reflection after completing Conversations.
This made me wonder: If we encourage learners to complete more Chats, would they be likelier to become committed learners? In other words: