I have long being sceptical of the costs and value delivered by our traditional methods of peer review. This is really on two fronts, firstly that the costs, where they have been estimated are extremely high, representing a multi-billion dollar subsidy by governments of the scholarly publishing industry. Secondly the value that is delivered through peer review, the critical analysis of claims, informed opinion on the quality of the experiments, is largely lost. At best it is wrapped up in the final version of the paper. At worst it is simply completely lost to the final end user. A part of this, which the more I think about the more I find bizarre is that the whole process is carried on under a shroud of secrecy. This means that as an end user, as I do not know who the peer reviewers are, and do not necessarily know what process has been followed or even the basis of the editorial decision to publish. As a result I have no means of assessing the quality of peer review for any given journal, let alone any specific paper.
Those of us who see this as a problem have a responsibility to provide credible and workable alternatives to traditional peer review. So far despite many ideas we haven’t, to be honest, had very much success. Post-publication commenting, open peer review, and Digg like voting mechanisms have been explored but have yet to have any large success in scholarly publishing. PLoS is leading the charge on presenting article level metrics for all of its papers, but these remain papers that have also been through a traditional peer review process. Very little that is both radical with respect to the decision and means of publishing and successful in getting traction amongst scientists has been seen as yet.
Out on the real web it has taken non-academics to demonstrate the truly radical when it comes to publication. Whatever you may think of the accuracy of Wikipedia in your specific area, and I know it has some weaknesses in several of mine, it is the first location that most people find, and the first location that most people look for, when searching for factual information on the web. Roderic Page put up some interesting statistics when he looked this week at the top hits for over 5000 thousand mammal names in Google. Wikipedia took the top spot 48% of the time and was in the top 10 in virtually every case (97%). If you want to place factual information on the web Wikipedia should be your first port of call. Anything else is largely a waste of your time and effort. This doesn’t incidentally mean that other sources are not worthwhile or have a place, but that people need to work with the assumption that people’s first landing point will be Wikipedia.
“But”, I hear you say, “how do we know whether we can trust a given Wikipedia article, or specific statements in it?”
The traditional answer has been to say you need to look in the logs, check the discussion page, and click back the profiles of the people who made specific edits. However this in inaccessible to many people, simply because they do not know how to process the information. Very few universities have an “Effective Use of Wikipedia 101” course. Mostly because very few people would be able to teach it.
So I was very interested in an article on Mashable about marking up and colouring Wikipedia text according to its “trustworthiness”. Andrew Su kindly pointed me in the direction of the group doing the work and their papers and presentations. The system they are using, which can be added to any MediaWiki installation measures two things, how long a specific piece of text has stayed in situ, and who either edited it, or left it in place. People who write long lasting edits get higher status, and this in turn promotes the text that they have “approved” by editing around but not changing.
This to me is very exciting because it provides extra value and information for both users and editors without requiring anyone to do any more work than install a plugin. The editors and writers simply continue working as they have. The user can access an immediate view of the trustworthiness of the article with a high level of granularity, essentially at the level of single statements. And most importantly the editor gets a metric, a number that is consistently calculated across all editors, that they can put on a CV. Editors are peer reviewers, they are doing review, on a constantly evolving and dynamic article that can both change in response to the outside world and also be continuously improved. Not only does the Wikipedia process capture most of the valuable aspects of traditional peer review, it jettisons many of the problems. But without some sort of reward it was always going to be difficult to get professional scientists to be active editors. Trust metrics could provide that reward.
Now there are many questions to ask about the calculation of this “karma” metric, should it be subject biased so we know that highly ranked editors have relevant expertise, or should it be general so as to discourage highly ranked editors from modifying text that is outside of their expertise? What should the mathematics behind it be? It will take time clearly for such metrics to be respected as a scholarly contribution, but equally I can see the ground shifting very rapidly towards a situation where a lack of engagement, a lack of interest in contributing to the publicly accessible store of knowledge, is seen as a serious negative on a CV. However this particular initiative pans out it is to me this is one of the first and most natural replacements for peer review that could be effective within dynamic documents, solving most of the central problems without requiring significant additional work.
I look forward to the day when I see CVs with a Wikipedia Karma Rank on them. If you happen to be applying for a job with me in the future, consider it a worthwhile thing to include.