Daniel Lemire's blog

, 2 min read

Toward author-centric science

Too many research papers in Computer Science are nonsense: they convey no worthy message. Yet they pass a Turing test of sort: at a glance, they are indistinguishable from interesting research papers. In fact, they are designed as nonsense from the beginning: the authors mimic the output of good research. The goal is to appear to be doing valued research. Some of these papers appear in top conferences, and even go on to be highly cited.

What is the difference between a Stephen King novel and the average horror manuscript? At first glance, probably not much. After all, it is all a matter of taste. Yet if you have read enough novels, you can recognize King’s mastery. Avid readers know who are the best writers and they stick with them. It is not sufficient to get a manuscript accepted by King’s publisher to get the attention of his readers. People first care for the author. Stephen King is the brand.

In science, we publish by venue, by journal and by topic. Yet if we followed specific researchers, the same way we follow specific writers, these researchers would have a strong incentive to produce interesting work. It would not be profitable to write many uninteresting research papers since nobody would follow your work.

The idea is not new nor original. Back in 2005, I urged researchers to setup RSS feeds for their publications. Arxiv recently made author feeds available. Creating RSS feeds is technically easy. We have no excuse.

Please: if you want me to read you, make it easy to receive notice whenever you publish a new research paper. If enough researchers do it, we might improve the quality science significantly. Why wouldn’t you want people to follow your work?

Further reading: Scholarly Communications must be Syndicated by Gideon Burton

Eating my own dog food: You can subscribe to my research papers by email or through a reader.