By Ng E-Jay
I attended the Institute of Policy Studies seminar entitled “Assessing the Rationality of Political Online Space” held on 11 Feb 2015. In this seminar, IPS presented quantitative research done on online political blogs in an attempt to assess characteristics such as journalistic objectivity, level of emotions, and level of partisanship. Regression analysis was carried out to identity and evaluate correlations between these characteristics.
Some interesting observations emerged out of this quantitative study. Firstly, blogs that had consistently strong political content were largely written by anonymous authors. It was hypothesized by some at the conference that this could be due to the lingering climate of fear that the government has instituted in the local political culture. Numerous defamation suits against well known bloggers like Alex Au and against former political candidates like Vincent Wijeysingha come readily to mind as factors that contribute to this climate of fear.
Secondly, there appeared to be no correlation between whether a blog was political or apolitical, and the level of emotional outburst displayed by the blog author. People who blogged exclusively about LGBT issues, for example, could display much more emotion than people who blogged about failed government policies. It was therefore hypothesized by some that it is the level of personal interest and involvement in the issue at hand that determines displays of emotional outburst, rather than merely whether the topic is political or apolitical. This is, of course, a very logical if not obvious deduction.
There appeared to be no correlation between whether a blogger was anonymous and his/her level of objectivity. However, bloggers who have made their identities known tend to write in a calmer fashion, perhaps because he/she has a reputation to protect. The majority (over 70 percent) of blogs tend to discuss both sides of the argument rather than merely give a one-sided account of the issue. Political blogs in particular tended to be two-sided in their approach, but of course this by no means applies to all political blogs.
Whilst these are certainly interesting findings, I personally find the use of a purely quantitative approach to be highly limited. I could tell from the presentation that this quantitative approach, utilizing keyword analysis and numerical classification algorithms, tended to leave out nuances in meaning and intent. To put it bluntly, the purely quantitative analysis misses the forest for the trees. It even makes mountains out of molehills because it tends to exaggerate certain characteristic about blog postings simply based on keyword repetition and usage, without taking into account context and intent.
As such, I was highly skeptical that the quantitative approach alone could accurately gauge the mood of the online political blogosphere.
There was some level of evasiveness when a question was posed on whether these findings were being used to aid the government. Conference organizers deflected the question by stating that findings are open to all, but admitted that raw data is shared with the government.
My personal conclusion from the conference was that for all the effort used in the quantitative analysis, perhaps a better, less costly, and more accurate approach would simply be to dedicate a small group of political analysts to read the more popular political blogs, twitter feeds, and facebook replies everyday and assess their content using human judgment.
Of course, logistic regression, correlation analysis, and keyword identification algorithms tend to be highly respected by the community, but being a mathematician myself I know when to shake my head and smile, and this is one such instance.
This article was first published at sgpolitics.net and is an exclusive republication for TOC.