User Research
/uxr57
a space for qualitative & quantitative user researchers to share insights, strategies, and stories.
I tried to make Devin.ai do user research data analysis.
It really is like trying to get a Jr Engineer to do something. You have to tell Devin what to do, HOW to do it, and what success looks like.
It's taught me a few things:
1. The prompt is everything
2. Create very clear examples of what "good" looks like
3. Don't expect a great output -- you still need to do the thinking
It really is like trying to get a Jr Engineer to do something. You have to tell Devin what to do, HOW to do it, and what success looks like.
It's taught me a few things:
1. The prompt is everything
2. Create very clear examples of what "good" looks like
3. Don't expect a great output -- you still need to do the thinking
one of us. one of us. one of us.
if they need anything to understand the product, it’s too complicated
agents will become the next wave of "users"
which means transitioning from "UX" to "AX (Agent Experience)"
different motivations. different constraints. different design space.
I wonder who will be the first "Agent Experience Researcher or Designer"
which means transitioning from "UX" to "AX (Agent Experience)"
different motivations. different constraints. different design space.
I wonder who will be the first "Agent Experience Researcher or Designer"
so many implications here — be careful what data you pay attention to
it may not be telling the story you think it is
it may not be telling the story you think it is
Observational user research can help you find these problems.
Find a customer. Shadow them. Understand the essence of the problem.
Brainstorm solutions and SHIP
Find a customer. Shadow them. Understand the essence of the problem.
Brainstorm solutions and SHIP
hacking on a new user research panel that leverages zkTLS to get proofs about participant data
just imagine if you could verify that a participant does X behaviour without breaking any privacy laws.
just imagine if you could replace a screener with a proof.
it would completely change how we do research.
forever.
just imagine if you could verify that a participant does X behaviour without breaking any privacy laws.
just imagine if you could replace a screener with a proof.
it would completely change how we do research.
forever.
the way you think about your product never survives first contact with users
I don't think we talk about danger cues enough
but how????
Really vulnerable and valuable reflections from Jayme…
UXRConf Summarization Podcast: https://notebooklm.google.com/notebook/d17cd9d5-5a05-4ee0-aabf-a8978ce9c843/audio
Condensing 10 hours of content into a 13 minute podcast
Condensing 10 hours of content into a 13 minute podcast
@markfishman pioneered the Loom feedback model that makes this easy.
record your experience end 2 end and share the loom link
record your experience end 2 end and share the loom link
feeling this so hard rn
I love offering user interview bounties on Bountycaster
- it’s quick
- participants are amazing
- get to support a farcaster native business
- it’s quick
- participants are amazing
- get to support a farcaster native business
try not to solve the wrong problem very well.
The more I hear consumer crypto product execs talk about their users, the more I think everyone's personas are basically a 1:1 match
👀 @openux
👀 @openux
I won't trust LLMs doing my qualitative analysis until I know for certain that it can actually "reason" better than I can.
Right now, LLM's are good for formatting/brainstorming screener questions, checking for bias, reframing and rewording interview questions...
Basically anything in the planning stage.
But for analysis and synthesis, what good is a fuzzy processor that is just predicting the next token, it can't actually pull out what is novel from my dataset. It's trained on what's known. It can't navigate the unknown.
Once it can reason, I'd happily give it the full task.
Right now, LLM's are good for formatting/brainstorming screener questions, checking for bias, reframing and rewording interview questions...
Basically anything in the planning stage.
But for analysis and synthesis, what good is a fuzzy processor that is just predicting the next token, it can't actually pull out what is novel from my dataset. It's trained on what's known. It can't navigate the unknown.
Once it can reason, I'd happily give it the full task.
tip: pay your participants while they're still on call with you
1) they'll love it (immediate rewards feel good)
2) it'll save you time because you don't have to remember it later
3) you'll feel good!
1) they'll love it (immediate rewards feel good)
2) it'll save you time because you don't have to remember it later
3) you'll feel good!