In this experiment, I turned one of those ideas into a user story using ChatGPT. And then visualize it with Google AI Studio. This was a fun, creative, and efficient way to go from an abstract concept to something visual and shareable. It helped me solidify the narrative of my idea and added an emotional layer that would be hard to achieve with text alone.

https://docs.google.com/document/d/18XwTqr9ZvonvdMIhD_S6v3WSDlswpmQUi29h1Nd9pcc/edit?usp=sharing

#aiexperimentlog #2Spotify #highsuccess

Building on the last experiment, this one was about remixing Two Concepts into a Unified Design Direction (with and without personal input). I asked ai to combine two of the stronger ideas into one new concept. Without the persona, the idea was generic and lacked emotional depth. Adding a persona made a huge difference: the revised output felt significantly more human-centered and aligned with the user journey.

https://docs.google.com/document/d/1H25BB_ejzukHnM8jRKCNloQpB8IzBeZGbkSmCbHLxeI/edit?usp=sharing

#aiexperimentlog #2Spotify #mediumsuccess

In this experiment, I asked ChatGPT to help me analyze last week's ideas, which I had color-marked as promising. It picked up on two really interesting questions — one about user-created filters and one about “taste overlap” with others — and broke them down into design challenges and opportunities. It felt like having a thoughtful design partner helping me prioritize what to explore next.

https://docs.google.com/document/d/1gVT1pjqJHXmA8l5eDrgKI6P5CifYo_3cKQUzVe4G45Q/edit?usp=sharing

#aiexperimentlog #2Spotify #highsuccess

In this experiment, I took a step back from AI-generated idea overflow and manually sorted ideas for personalized podcast recommendations. Clustering by theme helped me see hidden patterns—and realize just how much creative framing matters. I tagged four ideas:Taste Twin AI, Podcast Mood Board, “I’m Listening For…” Recommendations from Future You, to explore more possibilities.

https://docs.google.com/document/d/1y0oJmVuHYKsb7V642Pk5JfoYQO-rA8eWX8FcWi9-9xY/edit?tab=t.0

#aiexperimentlog #2Spotify #highsuccess

This experiment builds on my last one—I let AI generate a wide range of ideas. Same issue: too many broad results, making tighter prompts essential for quality. However, once I focused on a specific angle, AI provided surprisingly structured frameworks.

One standout: rethinking podcast recommendations through cognitive triggers and self-exploration—a fresh direction I hadn’t considered.

https://docs.google.com/document/d/148iE3JBI4wsg59mi5NdBiU3LX6BwxYzC_KijDxO-km0/edit?tab=t.0

#aiexperimentlog #2Spotify #mediumsuccess

I used ChatGPT to generate HMW questions from my Spotify personas. The results came quickly and were well-organized into categories. However, even with prompt limits, the output was overwhelming and lacked prioritization. My biggest challenge when using AI is narrowing down the scope of its generated results.

https://docs.google.com/document/d/1xgXSry_xFcIJLFMjxJAfXDxK-2Sd6KaB95S_eP9WXHA/edit?tab=t.0

#aiexperimentlog #2Spotify #mediumsuccess

In this experiment, I want AI to guide me to identify key insights from the persona which I can use for my next Ideation phase. The results were mostly expected, reinforcing the importance of a researcher’s own judgment. However, the AI introduced the P.A.T.H. framework, which I found surprisingly useful for analyzing my personas.

https://docs.google.com/document/d/1X-RO1_-bvqe93FV5Yn8en4HeT1WqjJbydIZnv9hboyE/edit?tab=t.0
#aiexperimentlog #2Spotify #mediumsuccess

In my last experiment, AI identified contradictions between two user interview transcripts, which I wanted to explore further. In this experiment, I used AI to incorporate these contradictions into persona building. Gemini synthesized two divergent user patterns into one persona and analyzed the conflict between the two users regarding recommendation usage, which may have been caused by differing levels of trust in the platform.

https://docs.google.com/document/d/12mO4GztKYNSuX1X3pwRAXjOomDWIm5_hBUy5obt0vPY/edit?tab=t.0
#aiexperimentlog #2Spotify #highsuccess

In this experiment, I explore whether ChatGPT can identify recurring themes or contradictions when analyzing two or more interview transcripts together—to support persona synthesis across user types. Additionally, I compare this generated result with a purely AI-generated persona, finding that actual transcripts lead to more accurate and less biased personas than those generated solely by AI without transcript input.

https://docs.google.com/document/d/1g0JxYc7SonxH4Zlof4ykMSzAx7D_iVnnKnvn8KSCMWc/edit?tab=t.0

#aiexperimentlog #2Spotify #highsuccess

I used ChatGPT’s “30-Minute Analysis Plan” to quickly analyze an interview transcript, exploring user behaviors and pain points in podcast discovery. The structured and efficient approach helped me identify overlooked details, such as brand trust, preferences for episode length, and frustrations with unclear platform descriptions. This was incredibly helpful for persona development!

https://docs.google.com/document/d/1ZbkZ5MSJZ8hT27DdW_jB4HILYE_jVW0rX_TO5hHxU1M/edit?tab=t.0

#aiexperimentlog #2Spotify #highsuccess