Is the arXiv a good way to get feedback on one's work?

There is not much of a "formal feedback" system associated with the arxiv. I say "not much" instead of "none" because the arxiv apparently does do some degree of automatic tracking of citations to its papers. For instance this arxiv submission contains a link to a MO post in which my paper is (briefly) mentioned. I had not seen that post before, so that was somewhat interesting. However, I don't know how this system works and it seems to be much less systematic than, say, what google scholar does. In particular, I have 25 arxiv preprints and the arxiv itself lists this kind of citation for very few of them, whereas google scholar lists much more.

In terms of an informal feedback system: yes, the arxiv works very well for that, in the following organic way: for many academic fields and subfields it is by far the one place to put your preprint in order to get the most (and the most interested) people to read it. This includes publication in most journals and the implications this has had on some academic fields are immense. E.g. I hear that in theoretical physics -- a fast-moving field in which it is apparently rare to look up a paper written much more than ten years ago -- pretty much everyone who is anyone uploads their preprints to the arxiv, and as a result theoretical physicists almost never go to the library anymore or look through actual journal papers: they don't need to. My field -- mathematics -- seems to be converging to this kind of phenomenon rather more slowly.

On the other hand, to expect feedback may be putting it a bit strongly. The volume of papers uploaded to the arxiv is fast and rapidly increasing. I just looked at the math.NT arxiv submissions, and last night 15 papers were uploaded. I am a number theorist with broad interests, and if these papers came at a rate of one a day or less, I would probably peruse about half of them. But the current volume forces me to be much more selective. The arxiv is great advertising, but all the advertising in the world doesn't guarantee that people will engage with your product rather than the sea of competing products (competing for their attention, anyway; they need not be competing with you in the academic sense).

In my experience, I most definitely get enough feedback from my arxiv submissions in order to justify uploading them (although there are arguments to be made for doing so even if you never hear directly from anyone about them). It happens that in the last three weeks or so I have uploaded four arxiv submissions. (Since I have 25 altogether, this is obviously a spike in the upload rate. Some other people do this too. Now that I think about it, from an advertising perspective it would probably be better not to do this.) Since then I have received comments on two of the four papers. The two papers that I haven't heard from yet are I think perfectly solid and interesting -- in fact, one of the two concerns the Combinatorial Nullstellensatz so probably has broader appeal than most papers I have written, and the other is a really substantial project that I did jointly with my PhD student -- so the fact that I've gotten no feedback about them seems to be mostly random.

In summary: yes, posting your papers on the arxiv is a great way to get feedback. Will it guarantee feedback? No, guaranteed feedback is exactly what you're buying (so to speak) when you submit to a journal. Other than that it seems impossible to guarantee. I would definitely submit to the arxiv and see what happens. If you hear nothing, then you might try sending a few emails to suspected experts which just point to your arxiv preprint. Having an arxiv preprint versus just enclosing a file adds a certain veneer of legitimacy.

The arXiv is a tool for establishing a presence, but it does not market or disseminate your results. It merely places them in a certain category and presents title, author, and sometimes abstract in a summary fashion, depending on how one uses it. Some researchers get RSS and email updates in their favorite areas about new arXiv submissions, but that should be considered small in number, and not likely to generate interest in your paper.

If you want feedback, you need to advertise your own work, on your webpage, at conferences, at society meetings, and other appropriate venues. You can prepare a short version (abstract or highlight only) and include the URL of the arXiv abstract. Use of arXiv does not indicate peer review, but as there is some endorsement system involved in arXiv submissions, there is also some cachet associated with having the URL.

When you have gotten some people interested in your work, they too can refer to the URL, and this can lead to more publicizing and hopefully direct feedback on your work. It can also lead to others posting their opinions on their blogs or elsewhere, which can be harder to track.

You need to find a way to advertise your work so that people find it on arXiv, use it, then eventually cite it and criticize it.

The main advantages of arXiv are:

• the publication timing. When you submit to some closed review journal or conference, you are months before an official decision. During this time (that can get long), you need a way to disclose properly your work;
• getting a larger audience. Everybody is not on her university's network, or some universities won't have access to all the journals. If your work is on the editor's site only, then more people than you might think will be blocked by the paywall. Furthermore, some people (I know some in Image Processing) will make monthly explorations of arXiv and publish some reading lists on-line, tghus giving you a larger audience.