3
$\begingroup$

SE.AI is a crowdsourcing platform comparable to Stackoverflow with an academic approach to discuss topics from Artificial Intelligence. It is open for everyone and the users can gain reputation by answering AI-related questions, for example about machine learning, algorithm theory and NLP. On the other side, the Arxiv website is a so called preprint server in which no discussion takes place, but the authors are uploading short papers of around 10 pages in the PDF format. The topics are the same like on SE.AI, that means it is about Artificial Intelligence. The problem with Arxiv is, that the website is restricted to the academic community.

Between SE.AI and Arxiv there is a gap. The postings in the online forum are short, in most cases only some paragraphs long, while the papers on Arxiv are the result of longer work. Become a member of SE.AI is easy, it is enough to press the “signup” button, but become an author in Arxiv is only possible for Phd students which have at least a 10 years career in Academia completed. The problem in Artificial Intelligence is, that most subjects are too complicated to discuss them without using the classical PDF format which contains of 20 pages and more. SE.AI is using URL references to papers on Arxiv, but it is not possible for SE.AI users to upload these papers.

So my question is: how to overcome the gap? Is it possible to lower the entry barrier in Arxiv, or is it possible to extend SE.AI with a preprint server for uploading papers?

| |
$\endgroup$
4
$\begingroup$

I think the short answer is to look for solutions outside either AI Stack Exchange or Arxiv, and publish in some other form.

Blogging is a reasonable compromise. It is a longer-form format than Q&A, less subject to rules to entry like Arxiv, and hosting services like Medium will get commonly-searched-for terms in front of interested people. Medium already has a strong presence of data science and reinforcement learning advice, with sample experiments and presentations of original ideas.

Is it possible to lower the entry barrier in Arxiv

That's definitely beyond the remit of the AI Stack Exchange community.

Also, in my opinion, there would be a large risk to the reputation of Arxiv. For every well-meaning researcher that would try hard to be thorough, read the existing literature, and actually have something to contribute, there will be many people with other motivations, other levels of conscientiousness and other levels of ability. It would drown out the useful articles, and there are already far more articles than it is possible to read.

is it possible to extend SE.AI with a preprint server for uploading papers?

Whilst it is technically possible, I'm not sure what it would add. For it to be meaningful, something has to happen to the "papers" after they are uploaded. Stack Exchange as it stands would allow for some comments/discussion about papers, and upvoting/downvoting them.

I think you may find yourself, out of the Stack Exchange community, in a small minority in wanting such a feature.

You could test this - it should be possible to create an Area 51 proposal which is about creating a beta site purely about reviewing semi-academic and pre-print academic "papers". Propose something such as the question text would be a synopsis plus a link to hosted PDF of the paper (perhaps uploaded/hosted by Stack Exchange), and answers would be critique of the paper using academic terminology - assessing it for presentation, technical merit, originality, potential impact. This could be framed as similar to Code Review Stack Exchange which made it out of Beta, so it seems workable.

There's a slight mismatch between Stack Exchange UI and what might work best for such pseudo-academic sharing and reviews. However, if you propose an idea like this, you can get some vague idea of support for it by responses in Area 51, even if they point out technical flaws but think the core idea has merit. Won't necessarily get what you want, but you'd get a sense of whether there are enough people interested that such a service could function. After all, there's no point you publishing your work if no-one is reading it.

Personally, I'd just blog about whatever you were researching. It's easy and the tools seem to fit what you want to do.

| |
$\endgroup$
  • $\begingroup$ I'm like you very pessimistic in overcoming the gap. As far as i can see, SE.AI doesn't even have enough resources to answer the 391 old questions, which have 0 answers right now. Peer reviewing pseudo academic paper additionally, is outside of the scope of this project. But thanks for your detailed answer, it make things more transparent. It is tragic, that on one hand it would be nice to have a huge amount of knowledge, which is provided by the crowd, but on the other hand, no resources are available which can do the job. $\endgroup$ – Manuel Rodriguez Sep 11 '18 at 15:08

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .