Specialisation or diversification in research

13th April 2016

My research focus has been something that I have thought about ever since completing my PhD. It is possible that I was even thinking about this during my PhD study. But during my studies the main goal of completing my thesis took a lot of brain space. But now, without the major project of a PhD thesis attracting my attention, I am a little more free to choose how I want to design my research career. So what should be the main focus of my research?

Specialise or Diversify?

What piqued my interest was an interview on the ABC Science Show with Professor Merlin Crossley, who is currently the Deputy Vice-Chancellor (Education) at UNSW. While the interview was predominately focused on Professor Crossely's research, there was a small part where the professor was asked about his views on the specialisation or diversification of education and research. It is Professor Crossley's view that people should specialise and develop a deep knowledge in one area. This idea really got me thinking because my own research is currently taking many different directions. So what does it mean to specialise?

The research performed in my PhD and current post-doc reside in very different areas of mathematical programming. In particular, my PhD focused on aviation applications of mathematical programming and my post-doc looks more into the computational aspects of mixed integer programming. So am I diversifying or am I developing a specialist knowledge? Maybe a better question is: does it hurt my career to either specialise or diversify?

Mathematical programming research topics

I would like to group my research interests into three broad categories: computational mixed integer programming, decomposition techniques and routing applications. This categorisation is not meant to form disjoint sets of research topics, but define some research areas to help guide the discussion in this blog.

This brings up the first point, if the categories are not disjoint, then is researching in each of these areas really diversifying? From looking at the literature, I believe that it is. The authors of papers in each of these categories rarely cross over into the others and when they do it generally involves applying the knowledge gained from their main area of focus. Now this is certainly not the rule and such statements can easily be described as generalisations. It is just an impression that I have.

As I write this I feel that I need to analyse my view on the categorisation of my research interests a little more. Is it necessary for applications, in particular routing applications, to be split from computational mixed integer programming and decomposition techniques? I am particularly interested in the application of mathematical programming, but I am also interested in the generalisation of solution algorithms. While the development of general and application specific algorithms may not be totally related, there are many examples of general algorithms that started their life as an implementation for a specific application. One very nice example of this is the Bienstock-Zuckerberg algorithm. So it is possible, and likely, that performing research on applications will produce results in general mathematical programming. Something very important to remember is that impact is only achieved in broader research areas if it is actively pursued, such as through publications and outreach activities.

I will round out this section by saying that given my current research ambitions, I will keep the three categories of research interests. I believe that it is valuable to look at applications and identify general approaches from the developed solution algorithms. Further, the ways to achieve impact in broader research communities, such as blogging and academic writing, are enjoyable for me.

Becoming an expert

As we progress through an academic career, it is expected that we develop an expert knowledge of an area. Is it possible to become an expert in any one area if you have broad research interests? As stated above, my research resides in three categories of mathematical programming topics. Developing an expert knowledge requires much time and dedication. So if there are many research topics that you are interested in, you could quickly become short on the time required to develop an expertise.

Becoming an expert in a research environment requires a strong knowledge of the current related literature. It can sometimes feel that there is more literature to read than one has the time. This is especially true if you are undetaking research in different areas that are not 100% related. For example, the published research of decomposition techniques and computational mathematical programming can easily end up in completely different journals. If you add different application areas to that, it is possible to become overwhelmed by the number of journals that you must have email subscriptions to in order to remain up-to-date.

So, in regards to developing an expert knowledge, the above reflection suggests that specialisation wins over diversification. However, that does not mean you can't become a specialist that combines different research ideas. Also, the problem of remaining up-to-date with literature is not completely avoided, but it does push your research into the niche category. If this is a desired career direction, then it comes with a commitment to continued and diverse reading.

Choosing a trajectory

A motivation of this blog is to identify whether it is possible to choose your own career trajectory and whether the choice is necessary. Taking the time to reflect on this topic, I feel that the answer should be yes to both of those questions. I think that how to choose is an equally important question. The answer could have a big impact on the outcome of your career.

I believe that you should make the choice of your research interests and career trajectory based upon what you enjoy. Success in research is output based and it difficult to achieve the desired level of output, in terms of quality and quantity, if you are not excited about what you are working on. You are the only person that will do the work, so it is important that you maintain motivation.

Bringing this back to my own experience, my desire to maintain the three research categories is based on what keeps me motivated. I find research on applications exciting because of the connection to the physical world. I also find the transfer of application specific knowledge to the general setting very interesting. In fact, I find this direction much more interesting than achieving solving performance improvement through the development of enhanced general algorithms. As such, I feel that I can achieve better research output by focusing on applications.

Final answer

What is the result of all of this? Do I think that it is better to specialise or diversify? What have I gained from this reflection?

My opinion is that Merlin Crossley is correct. By achieving a deep knowledge in an area, a person can push the boundaries of current research. By diversifying, it is not a clear where those boundaries lie. What is valuable is the collaboration of experts with different specialisations. This does not only mean collaboration with people from different disciplines, it could mean collaboration with people from within your broader research community. For example, research into large-scale optimisation problems from transportation applications may greatly benefit from the collaboration of mixed integer programming, decomposition techniques and applications experts.

What does this mean for me? I have been very fortunate to have completed a PhD with a focus on applications and now work in a post-doc position where I am developing my knowledge of computational mixed integer programming. At this point in time I have the ability to specialise in either of these two areas. I have realised that my greatest enjoyment comes from applying mixed integer programming techniques to applications. So I aim to specialise by focusing more on applications with the goal to apply my knowledge of general purpose mixed integer programming techniques.

What do you think about specialisation or diversification? I would love to hear your comments on Twitter.