I returned to the small study plan for Ideate - and got the idea to have Chat "regenerate" some past ideas. The result was good, but what really shone was the "roadmap" it asked if I wanted - kinda like a table, but somehow better. Check it out:

https://docs.google.com/document/d/1cEJFzfUIJhZua5wSz1XIrBWUCG8tWh0k1MM-AnGj06U/edit?usp=sharing

#aiexperimentlog #2spotify #highsuccess

Testing an idea for AI to "evaluate" ideas based on some cateogries - inspired by an experiment @siqijiang ran.

I think this was super helpful, especially so that I didn't have to sort through every idea myself. Kinda related to the "ranking" I did previously.

https://docs.google.com/document/d/1UNt78aXFqmViUrD1I4AeBGgn1fXvNrAcwklQ0fnznps/edit?usp=sharing

#aiexperimentlog #highsuccess #2spotify

Generating some HMW questions based on user needs! I felt this was more valuable when it expanded the questions and gave design opportunities as well.

https://docs.google.com/document/d/1jKG2BAfaoxmEqQcBz6ILSfuyaqY8yOGCfCmjjzggXnU/edit?usp=sharing

#aiexperimentlog #2spotify #highsuccess

The three tasks within this experiment: Generate, Cluster, Rank.

Which two do you think AI performed best at? Have you had different results?

https://docs.google.com/document/d/13PKbNn2LMAONOcWMZdrbV6urVHP--QESYN-Dy2dRJnI/edit?usp=sharing

#aiexperimentlog #2spotify #highsuccess

Worst part of GPT? Having to read it's long responses, in this quick-fire experiment, I wanted to see how it simplified our whole chat down into something skimmable.

https://docs.google.com/document/d/13MIxVyr6MitXqngAsbNItNBHunTobhAMg0swKbcAl3A/edit?usp=sharing

#aiexperimentlog #2spotify #highsuccess

Can ChatGPT regenerate ideas that are useful, actionable and innovative? That's what I tested in the following experiment - but I definitely confused GPT along the way - check it out here to see if you think it turned out to be a success in the end!

https://docs.google.com/document/d/1W6WA61WJxkxzQf_OvsJAbVNCGnlA4S3R369fIK-zQ8s/edit?usp=sharing

#aiexperimentlog #2spotify #medsuccess

💭 Welcome to the Ideate Phase! 💭

For my first experiment, I wanted to redeem the first experiment I did in my Define phase - trying to get ChatGPT to give me experiment ideas. (Which, kinda failed)

I refined my prompts based on what I learned last time- and got a lot better results! Would love to know what you think about these plans - especially if you have ideas on how to get weirder and wackier ideas!

https://docs.google.com/document/d/1XkefagvX_-GCPqGJZ0x2oj-uS22F-J1HEYN_79J3F80/edit?tab=t.0

#aiexperimentlog #2spotify #highsuccess

Following up the story generation - the natural next step is personas. Here's how chat dealt with developing four personas with (and without) a comprehensive user story.

https://docs.google.com/document/d/1dhUhAlIVXQUcnjeK0i93ID8nms1-UzFFm-FQfQZOHxY/edit?usp=sharing

#aiexperimentlog #2spotify #highsuccess

If you're in the user story generating business, I've found this two sentence template (from Laws of UX, I believe) to be helpful. Here's an experiment of me having GPT refine user stories it created for me to fit into this template. I found this super helpful and more informative than the "base" stories it gave me.

https://docs.google.com/document/d/1DnTo0sQjKGzohXoQdWNIEZLKN4Qosfbc8LDnn3FBeyM/edit?usp=sharing

#aiexperimentlog #2spotify #highsuccess

2-part experiment, part 2
(I'd never keep you waiting like the IG influencers do!)

MUCH higher success here - the TDLR is that you should write your own prompts, because ChatGPT does NOT know how to talk to itself (or, what you want without telling it!)

Came out of this experiment with sentiment analyzed, user needs/stories, AND the start of some ideation. (Stay in the phase plz, Chat!)
https://docs.google.com/document/d/1ugAqR-Xp9wCiLkTl8N2iXAqUWJTzOZ6cUuwravcZdnQ/edit?usp=sharing

#aiexperimentlog #2spotify #highsuccess