SP: So does user-centred design impact or restrict your vision for the site?
User-centred design is the only thing I think about--that's the critical aspect for designing Digg or Pownce, or any of the web sites I work on. All I care about is how people use the site currently, and how they will use it, depending on what I do.
SP: Do we need ideas any more then? Or should we just let our users tell us what we should be doing entirely?
No, no, our users don't tell us what we should be doing. Our users guide us to what the problems are. They'll sometimes make good suggestions, but usually they're suggesting a specific thing when they really mean a problem. An it's up to us to find solutions to the problems--like the Henry Ford maxim that I mentioned in my talk today: "If I had have asked my users, they would have told me they wanted a faster horse." Anticipating beyond what our users can imagine. Having the feel to make that logical jump to the next level. That's where the real genius is.
SP: In your presentation you spoke a lot about metrics, and about using information about how people really use your site to influence design. So what do you think about surveys? Should we be running them? What should we do with the results? And how much weight should we give the responses?
A certain type of person is going to fill out a survey, so automatically put that into your weighting. But surveys can be quite useful. We did one on the recommendation engine on Digg, and it resulted in some good data. But I wouldn't use surveys as your sole means of getting feedback from people. Surveys come across as quite tainted, so there's a lot of deciphering to be done with a survey. I think doing user testing, task-based analysis, and focus groups; bringing in a more targeted set of users is probably more beneficial. But it's more work than a survey too.
Get the whole article at SitePoint Design View.