He wants a "human" random distribution, not an uniform one. To see what numbers are picked most of the time.

The source shows he's also timing the pick, which I think is probably the more important dependent variable here. Doesn't matter what number you chose - it's how long it took you to find it, click it, and hit submit.

Indeed. The goal there is to see if people think about the choice, then second guess it before actually submitting.

Edit: Quickly skimming through the database, it also reveals "tab" loaders who open up a bunch of tabs and let them sit. (It couldn't possibly have taken someone 300s to make a decision!) Not the goal, and it does skew the data, but still interesting.

Mine will be slow due to trying to select a value on a slider with a well used G1 phone...

Don't use the data for timing purposes, that's more than, say 40 seconds?

How about not starting the timer until you see mouse activity on the page?

I hope he gathers data from different social groups. Would be interesting to prove that techie-types have a numerical bias different from marketing types.

Picks by country (based on IP geolocation) would also be interesting to see.

Just added this metric. Wish I had included it earlier, but hopefully there will be enough votes to get something interesting out of that.

It also seems he wants to see how different number-picking interfaces change that distribution. Try refreshing.