Tuesday, 29 May 2007

Computer Science in South Africa

In 2002, the South African government announced two things: a change in the university funding model which cut funding per computer science student by 40%, putting it below pure maths, let alone comparable subjects like physics or electrical engineering.

Around the same time, the government announced that it wanted to found an “ICT university”.

Clearly, the government thought the universities were doing a lousy job: it cut the funding in an area so important, the president thought there should be a dedicated university in the area.

Let’s look at publications and citations for the top 4 universities in the country, according to the Times Higher Education Supplement international ranking (available through Top Universities), from best down: Cape Town, Witwatersrand (Wits), Kwazulu Natal (merged from Natal and Durban-Westville) and Pretoria.

The following are all searches on the combined indexes, Science Citation Index Expanded, 1900-present; Social Sciences Citation Index, 1956-present; Arts & Humanities Citation Index, 1975-present (all searches done on 30 May 2007).

As you can see, all of the results show some level of research activity with some citations – but nothing brilliant (click on the pictures for larger versions).

Cape Town: 59 publications, 139 citations, 2.36 citations per item.

Wits: 46 publications, 58 citations, 1.26 citations per item.

Natal (includes Kwazulu Natal): 86 publications, 238 citations, 2.77 citations per item.

Pretoria: 84 publications, 151 citations, 1.80 citations per item.

The Natal results are skewed by the fact that there was a period when computer science was combined with geology, so these results aren’t an accurate basis for comparison and I will not consider these further. The others though present a consistent picture. The higher ranked universities have a higher publication count per academic (Pretoria has a much higher head count that Wits and Cape Town), but the general numbers are in approximately the same ballpark. The Wits figures should be considered in the light of the university having badly fumbled the ball on the management of the subject. The School of Computer Science there in recent years has collapsed from 12 academics to only 5, and recruiting is a shambles.

In any case, the government clearly didn’t think all this was so great. The ICT university idea however was replaced by creating a research institute, the Meraka Institute. Meraka has a few new people but is mostly comprised of the CSIR’s ICT division, Mikomtek, which has been rolled into Meraka.

Why, you may wonder, did the CSIR get this funding boost, rather than placing Meraka in a university? Clearly, if the universities are no good, the CSIR, and in particular, Mikomtek, must be a whole lot better. So let’s look at their numbers.

Mikomtek: 6 publications, 2 citations, 0.33 citations per item.

Meraka: 2 publications, 0 citations, 0.00 citations per item.

Now here is an interesting challenge for the reader: explain why the combination of Mikomtek and Meraka is so much better than any university in South Africa, to the extent that the CSIR has been made the sole custodian of this new bucket of money. Of course Meraka is in its infancy and could do better in future. It has after all only been going since May 2005. However, in 2 years, I would have thought that a well-funded institute would have recruited high fliers who would have published more than 2 papers that have made it into the top research indexes.

Perhaps the CSIR’s outputs are in other areas than publications. They generate a large number of press releases, for sure – and those are not listed in things like the science citation index. But that obviously counts for a lot in winning political support.

What about impacts on the economy which can’t be measured by publications?

The universities named here (and of course there are others) have produced thousands of computer science graduates between them. An organization which is not degree-granting like the CSIR has to do a lot to match that scale of contribution. One would expect many publications (oops) and other measurable impacts, such as commercial spin-outs.

For the latter point, let’s consider the case of commercializing the Internet.

In the 1990s, when the technology wasn’t commonplace, many small startups sprang up, touting the concept to business. The CSIR jumped in rather late and despite its massive government subsidy, it wasn’t able to compete. The top service provider in South Africa today, The Internet Solution, was one of those small startups, and was started by Wits computer science graduates.

Commercializing the internet is but one example; I believe it is up to the CSIR to make a case, rather than for me to tear them down, because the case for better funding for the universities is so clear. Universities produce graduates; the CSIR consumes resources. They need to demonstrate that they do so to useful effect.

Disclaimer: I applied for a position at Meraka in June 2006, and am still waiting for a definite response. I applied for a Wits position too, and am awaiting the promised feedback (no word for almost 3 months), so you can see that I am trying to curry favour.

For the record, here is my publication record by the same measure:

19 publications, only 1 citation – fair, not great. Not quite 3 times the entire output of Mikomtek plus Meraka combined. (My actual output is significantly higher than this – but to compare like with like, I am using the same measure throughout.) Then again, if I google on my name, I only get 28,500 hits; Meraka gets 160,000 [counted 14 June 2007] – those press releases really work. I should try that.