Research portfolios and research team composition: Lessons from swarm intelligence

The exploration of uncharted intellectual territory that is a fundamental aspect of cutting-edge research naturally creates numerous pitfalls and dead ends. For the researcher, the search for significant results thus becomes similar to the search for the needle in the proverbial haystack. Thus, in addition to due diligence — rigorous literature review, brainstorming at research team meetings, attending conferences to gain state-of-the art insights — a fair amount of luck is also helpful in moving a research program in the right direction. This is especially true early on in a researcher’s career; eventually, success tends to generate more success, and one gains a sixth sense about spotting promising research directions. This is true to such an extent that many seasoned researchers, when recounting their successes in hindsight, tend to forget the role that fortuitous decisions played in determining their career outcomes.

Career directions, of course, are by no means as clear-cut for young researchers in the midst of (or just fresh out of) Ph.D. studies. In this article, I propose a few rough insights that can be used in planning a scientific career. These insights are drawn from my own, partially successful flirtation with swarm intelligence about a decade ago. Swarm intelligence refers to a class of algorithms that are used to solve intractable optimization problems; their common feature is that they mimic the social behavior of animals, and use interactions among virtual individual agents to search through a mathematical space for a desirable solution. The most well-known techniques include particle swarm optimization (PSO) and ant colony optimization (ACO), although numerous variants named after different animals can also be found in the literature. These algorithms have been used to solve various practical problems, such as (in the case of my own work in 2005-2006) determining the best schemes for recycling water in large industrial plants. One of the keys to the successful use of swarm intelligence is to tune the algorithms to strike a balance between the “exploration” and “exploitation” characteristics of the virtual swarm. In the case of, say, a school of fish seeking food, “exploitation” refers to prioritizing immediately accessible food sources, while “exploration” refers to more speculative foraging behavior, which foregoes immediate gains in favor of the possibility of larger future rewards. What is clear from swarm intelligence literature is that excess emphasis on either extreme is detrimental; on the other hand, the best results are found by combining both characteristics.

How are such insights useful for career planning? Firstly, balancing exploration and exploitation is a principle that can be used for planning a research “portfolio,” i.e., a range of problems one works on simultaneously. On the one hand, any researcher wants to exploit research work where incremental successes will lead to a high likelihood of having significant publications within a time horizon of just a few years. Such successes are of course essential to establishing career momentum. On the other hand, some speculative research, with less immediate rewards, is also essential. In particular, such secondary research interests provide insurance to hedge against the possibility of one’s main research area becoming either obsolete or saturated. In my case, as a Ph.D. student working on life cycle analysis (LCA) at the turn of the last century, I initially developed a secondary interest in a related discipline called process integration (PI); PI has since become my primary research field, and accounts for roughly four-fifths of my career research output to date. A similar case (but on a much grander scale) is Francis Crick’s phenomenal, Nobel Prize-winning discovery of the structure of DNA, which was done as “side work” when he was a Ph.D. student in the 1950s. In fact, legend has it that Crick was advised by his Ph.D. supervisor to drop the DNA work and focus on his dissertation instead.

The second key lesson from swarm intelligence has to do with research team composition. For both virtual and real swarms, it often pays to have a mix of individuals with different exploration/exploitation characteristics. In the case of research teams, it also pays to have a good mix of temperaments. Again, this is something I have learned from experience. A case in point is my decade-long collaboration with Dominic Foo (now based at the University of Nottingham’s satellite campus near Kuala Lumpur), where one person’s creative flair provides the perfect complement to the other’s methodical rigor. We have since expanded our close-knit research team to include colleagues from the Philippines, Malaysia, India, Hungary and Taiwan. We have found that different individuals contribute different problem-solving approaches, perhaps based on cultural characteristics (for example, one of our colleagues, Santanu Bandyopadhyay from the Indian Institute of Technology Bombay, likes to derive mathematical proofs of optimality for many of the engineering tools we develop). Still others provide various unique skills, such as fault-free coding of computer models, eloquent wording of important concepts in joint research papers, detection of logical flaws in new ideas, and networking with industry partners and potential funding sources (there are too many names to list here, but they will surely recognize themselves if they read this).

In summary, my advice to achieving career success as a researcher, based on insights from swarm intelligence, are (a) to always invest a small portion of your time in a research interest which may not seem to promise immediate results, but may prove useful in the long run; and (b) to work in a team, preferably one where biases from different disciplines or cultural traits can actually be leveraged to generate ideas that would not be possible from a more homogeneous group. As a final note, I should point out that anyone familiar with swarm intelligence will know that they are heuristic algorithms, and thus do not provide a fool-proof guarantee of being able to find truly optimal solutions; however, they have empirically been shown to generally give very good results — which leads me to believe that advice such as this will generally be good for any young researcher to take.

* * *

Raymond R. Tan is a professor of chemical engineering, university fellow and current vice chancellor for research and innovation at De La Salle University. His main areas of research are process systems engineering and process integration. He received his BS and MS in chemical engineering and Ph.D. in mechanical engineering from De La Salle University, and is the author of more than 80 published and forthcoming articles in ISI-indexed journals in the fields of chemical, environmental and energy engineering. He currently has over 100 publications listed in Scopus with an h-index of 23. He is a member of the editorial board of the journal “Clean Technologies and Environmental Policy (Springer)” and is the editor of the book “Recent Advances in Sustainable Process Design and Optimization (World Scientific).” He is also the recipient of multiple awards from the National Academy of Science and Technology (NAST) and the National Research Council of the Philippines (NRCP) as well as commendations for three highly cited papers in Institution of Chemical Engineers (IChemE) journals. E-mail him at [email protected].