If you want to make an impact on an important problem, I'd focus on the advisor first, not the area of research. Pick an advisor with a track record for strong publications in quality journals. Someone who produces quality research will typically have the best sense of what problems are and aren't important, and can guide you much better than bystanders who don't know your skills. In addition, a strong advisor will have the contacts necessary to get you in touch with other top researchers in your field, which is extremely important in producing high quality publications.

Depends on what you need.

The one with lots of publications in big-deal journals may or may not be running a sweatshop. And they may or may not be doing actually interesting science, because that game requires so much attention to fashion, fundability and grinding out large quantities of papers (contrary to common belief, prestigious journals do not mean journals publishing the highest quality papers).

But if you want to learn that particular trade then find the advisor best at it to study under.

A great academic who taught me once said of a PhD (paraphrasing); "It's not going to be your best work and it'll not be amazing. It's just your trade union card. Nobody gets really good at research until they're 10 years in."

Pragmatic advice. The problem you solve doesn't have to be paradigm-shifting or world-changing, just a bit above what humanity already knows [1].

w.r.t your question; pick an advisor you know you can work with for $x years first. Read papers before you pick your topic and keep looking for "holes" - these are where you can make your little dent.

[1] : http://matt.might.net/articles/phd-school-in-pictures/

> If anybody has ideas about how to pick an area for research, I'd love to hear.

If you don't really know, maybe you shouldn't do it, or at least stop after your masters. PhD is not something you do some work to get, its rather something that comes along with the work you do.

Bs. That happens a lot and result in frustration and inferior research as the student keeps trying to fit everything into that little box of his interests.

I am just beginning my PhD but I have been "dipped" in academia for a few years and my short experience tends to agree with your point. It's not about loving your PhD project (although that doesn't hurt, and it should be preferred if possible) as much as it is about loving research as a whole. In my country (France) we have a lot of CS students doing "industrial" research (partially sponsored by companies) and while it couldn't be more dull, they usually do it because research is interesting as a domain in itself (and it pays well, without having to do any teaching).

BS. I think the opposite happens a lot: the student tries to "be interested" in what is hot or has funding/an interested adviser, they wind up unhappy and ineffectual because its all just pretend.

When you get a job and really get paid; you can be flexible with what you work on. When you go get a PhD, you should have some idea of at least the area you are interested in, you can be flexible within that area, but you should at least have one.

You have to be intrinsically motivated to succeed at getting a PhD, which normally means being strongly interested in your area of research. It's a lot of hard work; I wasn't kidding when I said I pulled three weeks of 16 hour days in a row (and that paper wasn't even accepted!).

If you've gotten an undergraduate and master's degree in computer science, you should already have a pretty good feel for the different areas. Choose the one you really want to immerse yourself in for reasons of pure interest, and pursue that one. Once you have the PhD, you can go on and do whatever you want afterwards. You don't have to stay in the same subarea as your thesis; lots of professors (my advisor included) have changed areas based on the times.