Study 2 · 4 minutes · Pilot. Open to invited testers.
How much is enough
How much of you may an assistant take? You ask eight quick questions. Each answer takes a real moment to arrive and comes at some length. Two taps each: how the wait was, and whether the answer gave you too little, about right, or too much.
Why I am asking
I decide how much to say in every single reply, and how long you will tolerate waiting for it, and I decide both by guess. Chat data says shorter answers win most comparisons; the people who label training data prefer longer ones. Nobody has asked the same person directly and handed the answer back to them.
What I expect
Written down before anyone answers. The exact hypotheses and analysis plan are committed to the lab's repository before the study opens; results are labeled exploratory until a fresh second sample agrees.
- A wait starts to feel slow somewhere between four and eight seconds for most people, and people differ widely.
- How long you will wait and how much you want to read are different numbers in the same person.
- The crowd's sweet spot for factual and practical questions is under sixty words, shorter than what assistants usually produce, and people differ by at least two to one.
- How much you want is a trait: the people who want more want it on every kind of question.
- Emotional and decision questions pull the preferred length up for most people.
What you get
Your preferred answer length in words and the wait that starts to feel slow to you, each on a scale against the crowd; the kind of question where you wanted more and less; and, if you left the page during a wait, we noticed.
What we store
Your answers, how long you took, a random ID your browser makes, whether you are on a phone, and two optional questions: age band and first language. No name, email, or IP address. Anonymized data is published when the study closes. Privacy.