How To: Find Papers to Read

Last time, we talked about how to read a paper. Most importantly, we noted that how you read depends in large part on what your intent is: you read differently if you’re trying to get a feel for a field as a whole than if you’re trying to specifically contextualize your work relative to specific other projects.

But to read a paper, you first need to find the paper! The class library is there to get you started on that, but it’s by no means exhaustive: it’s just meant to give you a clear starting point if you have no ideas whatsoever. Once you have some idea of the direction you want to go, though, what do you do next?

There are three general tactics that I recommend for finding research in a field.

Start With What You Know

Most every paper that you read is going to have some citations and a section on related work. The goal of these sections is to communicate to the reader where this paper sits in the broader community. Some readers might be able to draw those connections themselves, but for others, it is immensely beneficial to be able to see the relationships between a particular paper and the work that came before it.

Reading a paper without context is like seeing the solution to a math problem without knowing the original question. I have a paper in Learning @ Scale this year, for instance, about our online CS1 class. Its major emphasis is on the ability to deliver for-credit CS1 education at scale. To understand the value of that, it’s important to understand that most MOOCs are not credit-eligible, that credit eligibility brings with it accreditation requirements, and that other initiatives in the past have largely failed. If you don’t understand that context, then it isn’t as clear what the value of my paper is.

So, my first recommendation is: start with the related work of the paper you’ve already found and read. You don’t need to read every cited paper (that would take forever), but look through the paper and see where the citations are inserted. Chances are, you’ll find some interesting sentences, like “Research in the past suggests online courses lead to poorer outcomes and higher drop-out rates [1, 3, 5].” If you’re working in that space, you’ll want to know where those studies were conducted, what their specific outcomes were, and where there might be room for further exploration — so scroll down and find papers 1, 3, and 5 and jump to those next. Then, keep going recursively until you run out of sanity run out of time start to see some diminishing returns. Eventually, you’ll feel like you know enough to get started contributing, even if you haven’t read everything. No one ever reads everything.

As part of this process, keep an eye out for papers or names that appear a lot. These should guide you toward the biggest figures in the field. For example, if you read my work, you’ll find I cite Ken Koedinger, Beverly Woolf, Vincent Aleven, Chrisitna Conati, Roger Azevedo, Kurt Vanlehn, and Ido Roll a lot… and then if you read the other work I cite, or the work that cites me, you’ll find they cite those people a lot, too. So, that’ll lead you to big figures in the field that can serve as seminal anchors.

Use Google Scholar

If you’re not familiar with Google Scholar, you’re about to be. Google Scholar is great in that it allows some more intelligent filtering, organizes things a bit better, and connects to its own Google Scholar pages (like mine). Google Scholar is great if you haven’t yet found a paper in the area that you’re interested in; its search is a bit blunt, but it returns enough results that you can peruse a lot in a single, organized place.

Google Scholar supports two use cases that go beyond the approach I described above. One, it makes it far easier to find more recent papers and stay up to date with current work. Papers are often ranked and weighted by how many other papers have cited them, which is great for finding seminal papers, but citations are always backward-looking: papers published in the last couple years will never measure up in citation counts to papers from 5 or 10 years ago. So, Scholar will let you filter by this and find specifically papers published more recently about the topics you care about.

Second, if you follow the approach I described above, you’ll typically stay somewhat siloed within your community. If you start with one of my papers and trace it backward through the citations, you’ll read a good bit from AI in Education, the learning sciences, and cognitive science. However, lots of other places work on MOOCs as well: there are political implication that you’ll only find in PoliSci journals, there are evaluations of outcomes that are more contained within strict education conferences and journals, and there’s work done in the distance learning community more generally. Scholar doesn’t discriminate by community, so when you search for MOOCs, it’ll give you all of these. It’s a great way to make sure you’re not focusing too heavily on one portion of the community.

There are other services that work similarly and have even more meta-data to work on, like ResearchGate and Academia.edu, but I haven’t used them extensively — so try them out and let us know what you find!

Go to the Source

Finally, the third option is: go directly to the source. By source, I don’t mean the authors themselves, but rather the original places the work was published. There are conferences and journals for nearly everything, although they can sometimes be hard to find. Once you find one, though, you’ll have an archive of several years of papers within that general theme.

For example, imagine you’re interested in online education, but you don’t know where to start. One conference in that area is Learning @ Scale. So, you’d go to that web site or Google “Learning @ Scale 2017 proceedings”, and you’d find its table of contents. There you have a single page of dozens of papers within that theme.

Typically, these won’t link to full versions, but if you copy/paste a paper title into Google Scholar or some of the Georgia Tech library’s resources, you’ll very often get a full version. If you can’t, that’s ok: you can see the abstract and often a reference list, and those two things combined give you the most pay-off for your time anyway.

Which Should You Do?

I do all three of these things all the time, but where you should start differs in large part on where you are in your research. If you have no idea what you want to do, I’d recommend starting with some conferences or journals: just read through the titles and abstracts and see what you find interesting! If you already kind of know what you might want to look at, go with Scholar or trace the citation “breadcrumbs” of papers you already have.

But Don’t Take My Word For It…

If you want more, check out this article from Science about keeping up with the literature. Or, check out what some folks on Hacker News recommended. Finally, make sure to check out Google’s tips for using Scholar effectively.