Because
What defines computational sociology?

In my book, CSS has two big camps. One more empirically revolving oriented around prediction (typically using large data sets and machine learning: review article http://science.sciencemag.org/content/323/5915/721). The second one is more theoretically oriented (typically using agentbased computational model: review article https://www.jstor.org/stable/3069238). This is just a typology, the boundaries can be crossed. The languages you can use to these things vary greatly (R, python, java, netlogo, stata).

The coding language is epiphenomenal but people glom onto it because CSS is so hard to define. If you squint, I think it's basically multivariate statistics + datasets bigger than your runofthemill survey. But also a bunch of networks people seem to call themselves computational, as does basically anyone under 50 who studies culture quantitatively, so ¯\_(ツ)_/¯

I think this person meant estimator rather than algorithm. So it seems computation sociology is based around using nonstandard data collection techniques, "new" quantitative methods, and coding your own estimators (rather than using canned models)? Is that a fair assumption?

6a81 is translating algorithm (a computational term) into estimator (stats speak). They aren't the same thing. An algorithm produces a precise expected outcome. An estimator is about a model. To truly be computationally engaged, you could be doing either. Yes, noncanned models is one (estimator) approach. Another is developing your own algorithms (routines for computing specified outputs) for calculating measures. Your specification is on the modeling side, fac2's is on the measurement side. Both are prevalent among computational social science practitioners.
Think about a spatial modeler wanting to estimate distance decay effects. They could conceptualize, define and implement an algorithm to compute a new such measure. Then they could also develop an estimator (model) for that concept's effects on their outcomes of interest. Developing either requires computational capacities well beyond typing reg y, x. Neither is simply an if/then pair.

Also, fdc1, if you'd read the link you posted, you'd see that conditionals, etc. are potential components of algorithms, not algorithms themselves. Saying a conditional is an algorithm is like saying a lego piece and a set of instructions for building the Millennium Falcon from legos are the same thing.

Was on a search committee for this. You know it when you see it.
Of course you can do topic modeling, web scraping, etc in R, but teaching some Python was something we looked for.
Lots of very strong quantitative scholars applied, but the committee quickly boiled those out in favor of people doing the above mentioned types of things.Least informed SC of all time, sounds like.