Can a Google Algorithm Identify The Best Scientific Research?

It's always going to be (more than?) a bit hit and miss, but
maybe there are additional features that can be used for articles that are read on-line:

Average "time spent reading" paper (ignore very low times?)
References given in paper that are followed up
Weight citations according to the ratings of the paper from which the citation was made? Maybe the first two could be cross-correlated against the individual readers habits?

All I can say is...

No, a google algorithm can't identify the best scientific research. Even if a system of review was done on each article (such as movies in the International Movie Database), good articles will only be relevant to a select few. However, with this in mind, this system SHOULD be set up just to get people in the right direction. It must be understood (i.e. disclaimer) that the system isn't divine and is there to speed up the process of finding relevant articles which is a major weakness in all fields of study.

All I'm saying is you have this system, type in best article for whatever, it picks up some silly article that isn't the best (but probably okay), you scan the article for citations and good ideas/conclusions that could then be reinserted into the search engine in brackets or whatever, and the whole getting warmer approach. It won't be a one step algorithm but it sounds like it will require less steps.

You are probably right. This idea as originally presented is half baked at best, but if it could be put into operation there is a good chance that the need for improvements would become apparent, and there would be a structure in place to test the mods.