Risks and remedies for effective multidisciplinary research

Following my post about authorship networks, and the award of our multi-disciplinary AMR grant, I would like to share my thoughts on the structural risks of academic life to multidisciplinary research and some possible remedies.

Multidisciplinary research is of paramount importance. Many ‘big’ questions require a multidisciplinary approach: How to stop the spread of antimicrobial resistance? How to defeat cancer? How to measure gravity waves? How to land people on Mars?

Specifically, I will look at individualizing factors that  reward (or punish) academics or researchers as individuals rather than working together in multidisciplinary teams. Interestingly, these factors arise both from the traditional ‘scholarship’ model of academic life, and the modern ‘corporate/managerial’ model. For universities to best support multidisciplinary research it will require new thinking that both departs from the current managerial model, but also doesn’t go backwards to the traditional model.

Briefly, the individualizing factors from the traditional academic model are: the cult of individual brilliance; linear lists of research authors on publications; and research grants allowing only one PI. The individualizing factors from corporate university management are: individualized REF returns; workload planning models; performance targets; and strategic compulsory redundancies. And in the intersection between the traditional and corporate model: promotion. In detail then, each challenge followed by possible remedies:

Cult of individual brilliance. As academics, and as a society more generally, we are obsessed with the idea of brilliant individuals. Historically, ‘big’ research could be done by brilliant individuals. Charles Darwin, so we are told, got on a boat, looked at the birds in the Galapagos and came up with the theory of evolution. A similar story is told about  Einstein, trains and relativity. Sometimes we accept a great collaboration. Hardy and Ramanujan. Crick and Watson. Even among ‘normal’ (i.e. not world-changing) academics, humanities researchers speak of the ‘lone scholar’ while molecular biologists used to say ‘one grant, one post-doc, one protein’.  But actually these stories unravel. Take for example Crick and Watson. Most people now know that Rosalind Franklin played a pivital role in discovering the structure of DNA, and careful reading of the story reveal many other people who made critical contributions (e.g. Erwin Chargaff or Alec Stokes). Obsessions with big prizes do not help help: that Crick, Watsom and Wilkins got a big prize masks the fact that the discovery of the structure of DNA involved many contributions from many people.

How do we counter the obsession with individual brilliance? I think in truth almost all of us recognize that our work is not truly ‘our own’. We are all inspired by our teachers, our peers, our students, the people who wrote the brilliant paper we secretly wish we had written ourselves, or the people who wrote the mediocre paper who push us to do a better job. As long as we remember this, stop talking about big name prizes, and remove from our discourse the idea that ‘one person’ made some discovery – we can move forward. Importantly, we can have a big influence over the next generation in the way we teach. We have the opportunity to show to our students that the ‘big discoveries’ were mostly made by an interconnected collection of collaborators and rivals sharing ideas and being inspired by each other, not one or two heroes.

Linear lists of authors. This is a harder one to fix. Our research articles have a list of authors, and when we cite them, we typically only name the first on the list. Different fields have different conventions: some fields simply order from first to last in order of “importance of contribution”. In many sciences, the people who did the work get listed from the front, while the supervisors get listed from the back, so that first and last author have special prominence. In some fields authors are listed alphabetically. The problem with the normal convention in science is that hierarchies are established which give special prominence to two authors. How do you produce a fair author list for a complex, multidisciplinary study, where the findings reported result from genuinine collaboration between (say) three different laboratories?

So far, two main solutions have been employed. The first is to have joint first (or last) authorship. This works reasonably well insofar as a second author can indicate that their contribution is equal to the first. But it still suffers from the fact that one author is still listed first. The second solution has been to list author contributions. I remember when the idea was first touted it was said that a paper’s authors would become more like film credits than a linear list. But, rather than give informative detail, this has turned into a box-ticking excercise, with the contributions relegated to the bottom of the article.

In my previous post, I suggested a network view of authorship and contributions, rather than a linear list. To my mind, this solves many of the problems above. I received a number of questions about how the paper could be cited, and how credit would be quantified. To answer these, perhaps the network view could be enhanced in two ways. One possibility is to combine it with an alphabetic author list, so that there is no longer any hierarchy in that list. The second possibility – and I am not sure if I am sold on this – could be to (optionally) include a percentage contribution for each of the authors. This could satisfy the needs of those involved in bean-counting exercises (e.g. the REF).

Research grants allowing only one PI. Some grant organizations, for example RCUK, only allow a single PI on a grant. This means that where two or three people contribute equally to an application, their contributions cannot be equitably recognized. This situation is so easily remedied. The Wellcome Trust, for example, allows ‘Investigator Awards’ with two equal collaborators. Even the RCUK situation is inconsistent: two people from two different institutions can both be recognized as PIs, while if they were in the same institution one would have to be a co-I. We can call on all granting bodies to allow research grants to have more than one PI, listed as co-PIs, with joint and several responsibility for its delivery.

Individualized REF returns. For those outside British academia, you might be unaware of the REF (Research Excellence Framework): a process to apportion money between universities. University departments are assessed on the basis of their outputs (e.g. research papers that have already been peer reviewed and published) and then ranked. The REF can be criticized on many fronts; the problem with regards individualization is that departments need to decide which individual researchers to ‘submit’ to the REF and which not. Researchers submitted list their top four papers which are then given a number of ‘stars’ between 0 and 4. So with multidisciplinary papers from authors in the same department, this creates tension as to who ‘gets’ the papers for their REF return. And the stakes are high – an academic’s job is at risk if they are not submitted. Serious problem if two or three people in a department do their best work together and publish together.

Again, the remedy is simple. There is no need to individualize the return. The recent Stern review of the REF has suggested that different individuals in a department could return more or fewer papers – but it doesn’t go far enough! It would be just as easy to alter the REF so that university departments return their best papers – and these might be single or multi-disciplinary, involve one lab, two labs or many.

Workload planning models. Many universities in the UK are now implementing some kind of model of what work academics are doing, with points allocated for teaching, research and administrative duties. While on the one hand it is reasonable that tasks are shared in some equitable way, and on the other there is much to criticize about both the principle and practise of such models, here I will focus on the risks to multidisciplinary research that arise from individualization. Workload planning models are intrinsically individualizing. Each of us now has a ‘number’ on our heads – a proportion of a notional “100%” that we are supposed to be working – with risks of being too high (insufficient time for research) or too low (at best receiving an overload of tasks; at worst, redundancy). The problem is that we are judged as individuals – and need to ‘game’ the system as individuals.

Universities are complex, flat, structures, and a great deal of their success rely on acts of good will between colleagues (academic, technical and professional service). Historically, as academic staff, we are regularly doing favours for each other: standing in for tutorials, assessing and examing students, commenting on drafts of papers or grant applications and so forth. With the introduction of formal workload planning points, people start only doing those things that they get ‘points’ for. The helpful culture of saying ‘yes’ to your colleagues is at risk of being replaced with a culture of saying ‘no’.

Somewhere, somehow, these need to be pushed back against, so that we foster a collaborative environment. In the end, I think we need to argue against them for their weaknesses: it is impossible to compare teaching loads, research activity, administrative or leadership roles in a single system. Equity in teaching needs to be achieved through simpler measures of contact hours, not by risking collaborative environments through flawed metrics.

Performance targets don’t have to be individualizing but unfortunately frequently are. I am no fan of performance targets – but rather than critisize them outright, I will say here that where individuals are being told to be the PI on grants, or the lead author on papers, and that being a coI or middle author are being given less or no value, this will count against people carrying out collaborative work. There are many people, particularly those who have technology focussed laboratories (e.g. bioinformatics or mass spectrometry), who make very natural coIs on many grants – and that needs to be recognized and rewarded up to the highest levels. If there have to be performance targets at all, then at least these need to recognize collaborative working.

Strategic compulsory redundancies. There is nothing more individualizing than knowing that your job is on the line, and that you will be competing with your colleagues for a smaller number of jobs as part of a restructuring activity. I refer to ‘strategic’ redundancies as being a choice by the institution to cut jobs rather than to cut costs in other ways, e.g. infrastructure spend. (I contrast this with essential redundancies where the instutition as a whole is at risk and all spend is cut). Many institutions, even in the region where I live, have indulged in strategic redundancies. It is a disaster for multidisciplinary working if people are genuinely fearful about whether or not they will keep their jobs.

Quite simply, universities who wish to support multidisciplinary working should not indulge in strategic redundancies. This is a message the senior leadership teams need to hear and understand.

Promotion is probably the hardest of all to crack. Promotion is individual, whether seen through the traditional or managerial lens. Two people work very closely together on collaborative projects, and really neither would be able to be as successful in their research without the other. Could/should they be promoted together? (It is difficult to see how that might be implemented in any fair way). Can supporting the success of a colleague in of itself be part of a criterion for promotion?

On a different note, what about the technology-focussed academic I mentioned earlier (e.g. the bioinformatician or mass spectrometrist) who is coI on many grants, essential to many projects about the university, but rarely PI? Can this person be promoted on the basis of excellent collaborative venture? In order to support multidisciplinary research, universities need to recognize these types of contribution, and facilitate promotion to the highest levels on the basis of excellent collaborative, inter-disciplinary research.

The points I wish to make are not about any specific institution, but sector-wide. In the end, if we are to deliver on solving the big research questions, we need all of our employers to develop and foster environments that support and reward multidisciplinary research.





Authorship networks as an alternative to authorship lists

I am starting to think about different barriers to multidisciplinary research. One of the barriers is the traditional list of authors on journal research articles. The problem is that one has a linear list – and as a consequence the position on that list becomes hierarchical. Typically in our field, that means that being the first or final author on the list is over-valued, while other authorship locations are less valued. This then has an impact on jobs, promotions, etc.

Where research is genuinely multidisciplinary, this then becomes very problematic. Journals have tried to respond to this in various ways, including having joint first (or last) authorships and lists of author contributions (usually an afterthought at the end of the paper).

I propose here a radical alternative to an author list: an authorship network. This would replace the list of authors with a network (or graph) showing how the people have contributed to the work. Nodes on the graph could represent people, activities or grant codes. Edges could connect people to activities, people to grants (either as authors of the grant, or employed by the grant), people to each other (e.g. supervision relationships).

I have had a go at representing my most complex paper in this way. Here it is. Rectangular nodes are the authors. Rounded rectangles are the grants. Ovals are activities. Arrows between people link who is supervised by whom. Edges between people and grants represent grant authorship (blue) or employment (arrows). Edges from people to activities show who has done what, with thicker edges for the main contributors (i.e. Hiroki doing most of the modelling and Taku doing most of the experiments).

takahashi_et_alCompare the graph with the list of authors:

Hiroki Takahashi, Taku Oshima, Jon L. Hobman, Neil Doherty, Selina R. Clayton, Mudassar Iqbal, Philip J. Hill, Toru Tobe, Naotake Ogasawara, Shigehiko Kanaya, Dov J. Stekel.
I think the graph is far more informative. It is immediately clear that there are two main activities in experiments and modelling. It is clear that the BBSRC/JST is the central grant, while there is a contribution from the other BBSRC grant. Jon Hobman’s centrality to the research is also much clearer from the graphical view than appears from his position as third author on the list.
I’d appreciate some feedback and ideas.