Social computing, Computational Social Science and Sociology and Methodology in Computer Science part 2

Yesterday I wrote some conclusion and something of a summary on my thoughts. Today it was time to "kavla upp ärmarna" (roll up the sleeves) and get going - doing - *something*

Well lets call it research, as research in its truest sense, I suppose - literally searching and reading about essentially *everything* yet *nothing* and a sort of "throw something at the wall and see what sticks" type of method (not to be confused with throwing a strand of pasta at the wall and if it sticks it's ready to eat - not the single pasta strand you threw but the whole batch of pasta you presumably made (now that would be funny if you only cooked one strand of pasta)  - you can view it as a SAMPLE representing the WHOLE (population)).

See, research is about statistics, that is my ultimate conclusion perhaps.

Okay, enough with the shenanigans. It was time to work, work on that search query I suppose - so I included all sorts of Booleans and grouping and whatnot; but ultimately needed to go back and just look at individual words and terms. Any long search query is bound to be somewhat cumbersome, if you want to really understand what is going on, I think. But for demarcation|delimitation, it really is what you need.

At one point - I realized that the only 20 hits was actually all there was about this particular topic.

However, lets get back to the topic; I suppose, if I have anything cogent to say about this topic, which has not been said before (of course not): now: qualitative methods are old school - analog and manual. But, there is a way out of the misery, and that is by involving computers, naturally!

However, you still need to know what to do with this godly power of computation: first you can go easy with word frequencies and such... then it's time for: latent semantic analysis and latent dirichlet allocation - you can also stumble into stuff like probabilistic latent semantic analysis and latent semantic indexing.

Then it's time for machine learning, or if it was perhaps already included in the previous (may be the case): Supervised Machine Learning (SML) and get into that sweet Bayesian statistics.

Let the computer do the job, and sit back and enjoy. I guess. Well, you need to prepare the datasets and do the training and install a bunch of software and well, learn some new math and statistics including but not limited to linear algebra *gulp*. But other than that, just sit back and relax; the transistors will do the work from now on.

Well that would have been the case unless I had a manual method lined up for a literary review; which will likely need to obey certain rules. However; this meta study now has its subject or topic, which is all of above. I think it's a massive study but lets hone in on the particulars; which I believe will be related to the method (quantitative) to some extent and the analysis (content|thematic analysis).

Well it's a mixed methodology I suppose; but treating the data, which is qualitative? With a qualitative method (thematic analysis, using content analysis methods?) I guess you can go wrong here. There are divides which needs to be clarified I can *clearly* see. But I think that is what is interesting. Perhaps.

We will see, it looks more philosophical than anything else, but might do for a meta analysis I guess.

Social computing, Computational Social Science and Sociology and Methodology in Computer Science

These recent weeks I have been struggling with defining a field and defining a research topic; which should include research questions and topics. That also included some methodology in Computer Science and the related field Information Systems.

So lets start with what I have concluded so far.

The topic of social computing is a field or part of Computer Science which deals with social aspects of computing. I will not go into the exact definition here but I imagine it reads something like "social aspects of using computers and interacting with computers and computer information networks". I think that might be a decent starting point.

It should be quite "simple", yet as always in academia, there is a tendency to complicate things wherever possible so lets remember that. However, by simple, I mean, just take the words "social" and "computing", and it should entail the intersection of these two broader topics as well as where they intersect.

That would make the topic somewhat interdisciplinary. I think that can be a good thing, that I could delve into topics like social science and sociology, as well as perhaps some implementation of computer technology in for example networks.

Okay, so this leads me to the methodology part. As I have recently learned and pondered is that computer science traditionally could be viewed in the positivist tradition. That would make sense, as computers are quite quantitative in nature and therefore would oblige by a classical scientific approach.

However, when working with social aspects, qualitative aspects also become important. Now, one can go either way: purely quantitative or qualitative but I think a mix might be nice. I'm thinking of a methodology where both aspects are being taken into account. (I will not go deeper into this at this point, but there's a case to make for choosing a mixed methodology in this particular case -- see below for somewhat of an example).

When I researched this and read about it, what crossed my mind was the meta level of this, or rather, the "computer science" angle, rather than just using some computer technology in a qualitative method. I guess it's wishful thinking but if quantitative method could be applied directly to "social data" - perhaps something would come out of it.

And it appears that this has been done, especially in the field of computational sociology, where a lot of interesting computer based methods are being used.

It also made me realize that another interest of mine; which I wasn't sure was quantitative, actually was -  text analysis. However, I think it holds, or can hold, some qualitative aspects as well.

Well, this might too simple, or it is exactly what science should be about - connecting different areas which haven't been connected before (well you'll most likely won't be the first, but great minds think alike a what not)...

I could also see that one uses quantitative method for "exploratory" purposes, and then, perhaps defines the research problem and topic a bit further. Then adding a qualitative "measurement", and by combining the two, getting further in the analysis, than just using either of the two.

For instance: get the word count of a certain topic in a certain text or group of texts, or get the top 10 word counts in a certain text  or group of texts. Lets say that word is something crucial to whatever is being studied, then it would make sense to learn more about this word and its meaning in the context, rather than just running the word frequency function and stating that this or these word/s are prevalent.

Well, this is what I've been thinking about and this is the preliminary conclusions I guess. Now I just need to pick a relevant "sub topic" and go ahead with my methodology... Unless it seems to be the case that I will need to use a specific methodology.

However, what I have described here, will be part of my "theory", I suppose. Or what I except to find. This will the be considered somewhat deductive, although I suspect there may have been some moments of "induction", as well. Or rather, my chosen sub topic, may or may not have support for what I have just posited.