2017 August - Leader

Eating Chihuahuas – The OR Perspective


One of the recurring themes of my presidency so far has been the diversity of OR and the opportunities and challenges this brings. My excuse for bringing it up again is that every OR event or conversation seems to showcase diversity (and I want to share my excitement at this) but also to re-surface one of its key challenges, which I shall come to. I have just spent three enjoyable days at the annual conference of the Portuguese OR Society, APDIO, where I was invited to give a plenary talk on the UK’s experience with Pro Bono OR.

It seemed a good idea to sample one of the parallel sessions, and I chose ‘Metaheuristics 1’. To be honest, I didn’t understand much of the detail of the four talks, not (just) because metaheuristics is not my strong point, but because my Portuguese doesn’t run much beyond “Desculpe, eu não falo português”. 

Although the talks were in Portuguese, the slides and abstracts were in English. And that’s when it struck me that the topics being addressed within one small technique area – railway networks; diabetic foot; beer brewing; ambulance location – represented an extraordinary variety. APDIO being a relatively small society, only relatively few techniques were represented. In the last 12 months I have attended a EURO conference, The OR Society annual conference, the Beale lecture, the new ‘Mathematics of OR’ conference, the ‘Young 2 OR’ conference, the Annual Analytics Summit, and The OR Society AGM, and I have spoken to many OR people at these events and elsewhere.

The number of techniques currently in active use by OR people is enormous. And to that, we add the variety of approaches in the skillset. My own talk on UK Pro Bono had exampled three projects – simulation to help get the most out of call-centre volunteers; analysis of donor data to help boost fundraising; facilitation to enable a vulnerable organisation to decide strategic direction in 2 hours.

Of these, one used a recognised OR modelling technique, one used statistical methods to produce actionable insights, and one used a process tailored to the specific circumstances, to enable problem-owners to shape their own solutions.Pro Bono projects are often (not always!) technically trivial, but important and challenging to the problem owners, and often with local conditions (constraints, personalities, stakeholders etc) that require sensitivity, skill and straight thinking to handle.

This is true of many of the problems that practitioners (especially in-house practitioners) tackle. OR also addresses plenty of problems where it is the technical challenge that dominates. And then there are problems that require every skill OR people can muster, where the technical difficulty is only one part of the larger challenge of implementing change that will address the specific problem.

Given this variety – industry, problem, technique, skill, scale, nature of challenge, and more that I haven’t mentioned – it is not surprising that ‘diversity’ keeps popping up as a theme. This diversity should be celebrated. It is inspiring and amazing that OR can and does improve the world in so many different ways. It gives us opportunities for practice and research; it gives us access to new ideas to inject into our own work; it gives us a profession we can be proud of.

Of course OR people are problem-focused, so we usually skip the celebration and go straight to the accompanying downside, the key challenge I mentioned at the start: how do you define something like OR? The tighter the definition (for example definition by technique), the easier it is to find an OR professional who is doing something outside that, and/or a non-OR professional who is doing exactly that. The looser the definition, the more it needs straplines or additional information to convey the unique essence of OR. Do we need to define OR? Well, no, certainly not for drawing strict boundaries to keep the outside out and the inside in. But we need some way of explaining to potential customers or funders what value we can add, of clarifying what our events and publications cover, of enlightening potential recruits about their options, of pointing OR people in suitable directions for professional development.

And because we do have, in OR, a distinctive and important discipline, it is important to maintain our shared confidence in what makes it distinctive and valuable.This is where the chihuahuas come in. You may have seen Karen Zack’s “Animal or Food” picture challenges (“puppy or bagel”, “parrot or guacamole”); and the response by Artificial Intelligence experts. Computer recognition has developed very fast, as viewers of Andrew Blake’s Blackett Lecture last November will know. This example of “chihuahua or chocolate chip muffin” https://blog.cloudsight.ai/chihuahua-or-muffin-1bdf02ec1680 shows just how good it can get.

Exciting? Scary? Either way, it is clear that Artificial Intelligence (incorporating classic OR techniques) has powerful methods for sorting things that are in many ways similar. So why not use it on ourselves: take OR abstracts, teaching materials, project write-ups, funding bids, and let machine learning work out for us whether something is or isn’t OR? Would that work or would it expose the limits of machine learning, or of our training datasets? I leave that as a challenge for the appropriately-skilled reader. 

In the meantime, here is the approach I have found the most useful whilst meeting the Presidential requirement to talk to a lot of people about OR.

Step 1: Recite the most open definition of OR that I feel comfortable with (e.g. “Helping organisations become more effective using scientific approaches…[breath]… i.e. grounded in evidence; rigorous, consistent, valid, logical; using models/algorithms/hypotheses to explore”). 

Step 2: Draw out of my bag the latest copy of Impact, and hand it over, saying “For example…”. 

Step 3 almost always follows: an engaging conversation about what OR people can do. Which is, after all, far more interesting than the question of what OR is.