Comment by grkvlt | Hacker News Reader

Comment by grkvlt | Hacker News Reader

Parent Full thread

Don't citation indexes work pretty much exactly like Google's PageRank, with cites taking the place of href links? A parent comment seems to suggest the same thing, so it looks like the ranking is gameable.

In practice I understand this happens via citation-swapping (similar to link swapping, or paid links in blogs) and poor-quality journals accepting articles for a fee, or without review (much like paid content, or content farms) and papers that cite other articles just for the increase in rankings (similar to link farms) so the analogy to SEO (in its various hats) seems to fit.

This would suggest that the only way to fix the 'broken' citation ranking system is the same as Google, and bolt on empirically determined heuristics above the pure mathematical model of PageRank, and black-list or penalise certain journals or types of citation?

> and bolt on empirically determined heuristics above the pure mathematical model of PageRank

The impact of Google doing this is, in financial terms, probably much larger than if the same type of approach would be applied in the academic situation. So if Google is allowed to do this, why can't we allow the same approach for research? Apparently, the system works quite well.

Honestly, funding a research team to develop a science ranking engine that was both being continually developed to thwart gaming and effectively independent of university and science-funding politics would be a truly excellent use of YCombinator's money.

Indeed. Simple link counting based search algorithms got gamed a lot in the past, by regular website operators and most heavily by spammers. That's partially why the current search algorithms are much more clever and secret than just using the Page rank. Human raters and machine learning off the data they generate are used as well. This helps to make the rules more uncertain and nasty SEO tricks much harder to discover and use.

Compare this to the situation in rating of academic research, where the papers are rated by the community of authors itself and the ranking is essentially a weighted counting of references. Of course it leads to lots of self-serving spam.

If the idea of independent raters (and possibly machine learning on top of it) could be implemented for academic papers as well, that would be a big improvement, I think. Academia has got a lot to learn from search engines.

Reply on news.ycombinator.com