Alessio Guglielmi's Research and Teaching / Politics
There are some possibilities of abuse in anonymous refereeing, because of the large irresponsibility of the referee towards the author. Sometimes I receive careless and, in a few cases, dishonest reviews. I know it happens to others, so I wonder whether there is anything we can do to limit this phenomenon.
I believe that referees should accept that their reviews are published on the web site of the author, and possibly answered in public (like book critics do for books, say). Alternatively, or in addition, referees could sign their reviews. The norm should not be too strict: one can imagine circumstances where disclosure is undesirable, and anonymity should in those cases be acceptable.
Publishing reviews serves two purposes: 1) careful, positive reviews can persuade others to read the paper; this way the work of the referee has more value; 2) negative reviews can be answered and commented in public, and this, of course, places some responsibility on the referee.
There is a danger: referees might be induced to be too merciful, and obviously this is not good because science does need high standards. It is a matter of establishing a new culture and associated ethics, therefore of finding the right balance. If more and more people adhere to principles of 'responsible refereeing', I believe that we will all benefit from a healthier reviewing.
Therefore I decided to sign my reviews and to give explicit permission to the authors of the refereed work to publish and comment the reviews they get from me.
All that said, I believe that this business of formally refereeing papers prior to publication is a thing of the past, when access to communication channels was limited and expensive. The internet has changed the picture. In my opinion this is the way it will work when the inevitable resistance is overcome: a researcher puts a paper on the web, then a colleague reads it and finds it good and writes a comment, or suggests changes, etc. Others find the paper useful and correct, and will cite it and comment it, and so on. The bad stuff will be ignored and therefore automatically filtered out and forgotten.
Bertrand Meyer, Aaron Sloman and Doron Zeilberger say it better.
Many of the ideas above are being implemented by a new publishing company called ScienceOpen. I have joined their Editorial Board.
I believe that there is a problem with the conference system in computer science, at least for the part of it that concerns theoretical computer science.
Theoretical computer science is a branch of mathematics that happens to be practised, mostly, in computer science institutions. Contrary to the tradition in mathematics, most theoretical computer scientists publish their work in conference proceedings, and only to a much lesser extent in journals. This practice is strongly encouraged by the fact that conference publications count towards promotion and tenure.
There are many good aspects to conferences, and especially, of course, their role in facilitating communication and circulation of ideas. However, the following negative aspects, in my opinion, largely outweigh the positive ones:
These problems would not be too serious if conferences were only a means of exchanging novel and perhaps still immature ideas. However, hiring and career advancement depend on conferences, and this confers them disproportionate impact. We are building a system without solid foundations for the production and vetting of scientific results. Of course, good institutions and good scientists still produce good science, and good conferences have good papers, but what about the average quality? In my opinion, it is pretty low in computer science, and the conference system has a big role in that.
I believe that there is also a subtler and worse problem associated with the peculiar conference system of theoretical computer science. There are many conferences, and we have developed a very competitive culture, to the point that you cannot be a top gun if you don't have papers at each and every top conference. So, we have some of the most brilliant minds constantly working at small problems, those that fit the conference format and that can be ready on time for the next competition. The same do the less brilliant, or those who work in unfashionable areas, and who compete for second-tier conferences. All these people would probably do better and would be more creative if they felt free to work on less incremental research.
It is very sad that the first thing we have to tell our students is that academic life is a rat race, and that either they publish or they perish.
So, what can we do? The origin of the problem is not in conferences, of course, but in making conference publications count for career. Many agree with this, but then many also say that the problem is in the system as a whole, or in the funding mechanisms, or in the culture, and so on. Many also say that everything is going to change soon anyway, because the internet changes the economics of publication, and so the traditional ways are all doomed. Whatever. The point is that, at least for now and for some of us, refusing conferences can be not only socially responsible and an exemplary behaviour, but even advantageous to our research. It is a pity that some of the very best among us do not even notice this problem, probably because they are very 'successful', and therefore gratified, in the present system.
Since 1995, I refuse to be involved in conferences, except when vulnerable colleagues might be penalised by my doing so. More in detail, I do not send my papers to conferences and I refuse invitations to speak at them and to referee for them. This applies only, of course, to conferences with published proceedings, which are the conferences that play a role in career advancement. The conferences where people only go to meet colleagues and exchange ideas are perfectly acceptable. This policy has slowed down my career, but it has not killed me and it has served me well, because I could concentrate on tough problems and then develop an ambitious and broad research program.
Of course, as everybody, I have responsibilities towards my employers and colleagues, and so I have to correct possible excesses. I do this as follows.
Communication of scientific results (so that they have an 'impact', as the managers say) is an important part of research. My employer, and eventually the tax payer, have the right to see their money get into more communication than just journal papers. To compensate for my lack of participation in conferences, I organise several meetings in my field, such as workshops and summer schools, and I take great care in communicating my findings as effectively as I can.
I also have to be careful not to stifle the career of my junior colleagues, especially PhD students. For them, I make an exception to my rule of not sending papers to conferences: I do occasionally send conferences papers that are coauthored by colleagues who need a permanent job, provided that they present the paper. Of course, this should not be a free ride, because the conference system costs in terms of refereeing. Therefore, the deal is that my colleagues whose work has been submitted to a conference give back the same amount of refereeing that they receive. This seems to me a fair compromise to everybody.
I am interested in hearing comments (both positive and negative) about my position on the matter: just drop me an email.
These are some pieces of the debate that is going on in theoretical computer science about the conference issue mentioned above: