SW23 Keynote
Professor Susan Howick, Professor of management science and Vice-Dean (Academic) for Strathclyde Business School, delivered a keynote at this year’s Simulation Workshop (SW23) which explained clearly, how and why methodologies could and should be mixed in simulation work.
In consideration of mixed methods, there are publications available, which discuss the frameworks for mixing OR methods in general and specific to simulation. The theoretical considerations behind using them are discussed in detail. There are many case studies to be found too which explain exactly, how methods can be mixed. (General Mixed Method, by Howick and Ackerman 2011 and Hybrid Simulation, Brailsford et al 2019).
Methods can be mixed in parallel or sequentially. When a parallel approach is selected or applied independently “comparisons can then be drawn between the methods used”. This is a form of “enrichment” where a primary method is enriched by the application of some elements of other methods.
The sequential approach is one where methods are applied one after the other. They can be seen in total, as just a method or as elements of methods that can be combined to form a new method. In both instances of method mixing, it would be possible to see if interactions between methods occurred during simulation.
A lot of interest has recently been shown in mixing System dynamic (SD) and Agent Based methodologies (ABM), there are benefits in adopting this approach, and Professor Howick gave several reasons for this.
Information can easily flow from SD to ABM and vice versa. As an illustration, she presented a slide that showed how such an approach could work, for example, in monitoring stock levels by defining agent-specific stock variables – agent-state variables affect flows too. Stock levels can also define the behaviours of individual agents and the behaviours of agents can affect flows.
Professor Howick spoke about the qualities and skills that simulation modellers should exhibit and be familiar with before attempting mixed methodology problem solving.
Modellers should be prepared to be, “open to working with other modellers and disciplines”.
They should speak a common language for effective integration, “the main reason projects fail, is the inability to work together and speak together because they (the modelling team) aren't communicating. It's a skill that is not particularly widespread”.
Additionally, the modelling teams should challenge each other and understand other simulations, methods and approaches, teams should trust/respect each other. Such modellers should “see the big picture” and recognise the frameworks that can be used to assist in the choice of mixing methodologies.
With any project, and not just projects requiring mixed methodologies, there should be a leader/integrator, “who should possess good team leader skills, a broad understanding of different simulations and methods and have the ability to act as a translator to facilitate conversation between modellers”.
Mixing methodologies is still in its infancy with few practitioners practicing it. The value is by having these methodologies talk to one another. The output of the optimisation wouldn't be there unless the simulation was there.