Can AI help us be more reflective?

See full project on Github.

Hypotheses and Testing

Through user testing and iterative design, we aimed to address the key hypotheses: AI can enhance journaling by reducing effort and increasing engagement.

Feedback from users revealed mixed responses. While many appreciated the ease of recording and reviewing entries, others found the integration between voice and text disconnected.

“I liked the concept!! Voice memo + written entry is awesome, but I felt like the text and voice didn’t feel connected.”

To address this, we improved the visual signifiers for recording and transcription, such as adding live transcription indicators and timestamps.

During our usability testing, navigation and design elements were areas of frequent feedback. Many users struggled with the app's flow and visual cues:

“It wasn’t super clear that the phone was recording visually. Discard was full red the whole time, which distracted me from save.”

“Flow was confusing without any entries at first; might be more intuitive to have some kind of onboarding.”

We made multiple changes, such as designing a step-by-step onboarding experience to introduce new users to the app. We also updated visual cues for recording, such as replacing the red discard button with a less distracting icon and adding a pulsating indicator for active recording.

User Feedback Highlights

To better understand user experiences, we conducted multiple rounds of usability testing. Below are some key insights:

• "Widgets felt bulky and unclear." Streamlined the widget interface to prioritise usability and clarity. "

• Kept accidentally hitting add and couldn’t exit without submitting." Added a cancel option to the entry creation process.

• "I didn’t know why text and voice appear at the same time." Enhanced the synchronisation between voice and text, with clearer visual pairing.

• "Sorting is intuitive, but overall organisation is confusing." Introduced filters and sorting tools for easier navigation of entries.

• "I’d like more control over summaries and sharing." Added options for user-driven summaries and future plans for sharing capabilities.

Ethical Considerations

1. Resurfacing Negative or Triggering Content

Echo is designed to provide emotional insights while respecting the sensitive nature of journaling. To prevent distress:

• Sensitivity filters flag potentially triggering content.

• Users can opt into summaries for sensitive entries, retaining control over their experience.

• Planned feedback mechanisms will allow users to report inappropriate AI behavior.

2. Inclusivity for Linguistic and Cultural Nuances

Users noted challenges in transcription accuracy for informal language and underrepresented dialects. For example:

“I don’t know if this would work well with Cantonese slang or Singlish."

To address this:

• Future plans include fine-tuning AI models with diverse datasets.

• User feedback mechanisms will ensure continuous learning and improvement.

3. Reliability and Data Integrity

Local storage posed a significant risk of data loss:

Customer Quote: “If my phone breaks, do I lose everything? That would be so frustrating.”

Transitioned to cloud-based storage for secure backups and cross-device access. Manual export options are planned to give users full control over their data.

Conclusion

“I think this is cool, but it depends on how much user control. I’d like it more if I could do it myself and maybe had some sharing aspect.”

Through Echo, we were able to explore how generative AI can empower users to reflect, stay grounded, and capture their stories. However, the question remains whether users would embrace Echo Journal's AI-driven tooling to summarise and synthesise something as personal to them as their memories.