I recently saw my first silent movie: The Passion of Joan of Arc. Much to my surprise, I loved it. I can’t wait to see another (I’ve got my eye on Nosferatu). I loved just about everything about the Passion of Joan of Arc and I didn’t miss the lack of dialogue. The movie translated just a few of the lines that were “spoken” during the movie. Much of the dialogue was left for the viewer to “hear” by reading lips and body language. The acting and camera work conveyed the emotion of Joan of Arc’s short life and trial, and it packed an incredible punch (Roger Ebert explains why this movie is a masterpiece much better than I can).
After seeing the film, I reflected on movies and technology. Movies are not just a product of technology but are also shaped by technology. Each time that movie technology changes, the movie experience fundamentally changes. Because special effects and dialogue are so easy to put in movies these days, there is too much CGI and talking to have an emotionally moving moviegoing experience most of the time. Most movies fail to convey what they really should be conveying, and as a result are flat and uninspiring (There are plenty of exceptions, and a quick glance at my Netflix history includes Tully, Gone Baby Gone, Lars and the Real Girl, Bella, and Juno).
The same can be said about operations research. We work in a technology-driven field, where we rely on software to do much of our work for us. Sometimes we rely on software too much and on good modeling too little. A IIE blog entry writes about blindly using software as a quick fix. When computing power wasn’t very powerful, making a tight, efficient formulation was necessary for finding optimal solutions. I hope we haven’t lost some of the art of OR.
This past week, I have been working on a research problem that is formulated as an integer program. I was curious to see what the optimal solutions look like, so I popped it into CPLEX. After letting CPLEX churn away on it for more than 72 hours, I felt a little guilty that I didn’t spend more time to make an elegant and efficient IP formulation. There may be a great theoretical research problem to work on, but alas, I know that it was just laziness. Just because computing is cheap and easy doesn’t mean that I shouldn’t pay attention to theoretical aspects. Although to be fair, I thought there was a good chance that CPLEX would very quickly find an optimal solution, which offered little incentive to consider various formulations. We don’t want to lose sight of important theoretical contributions just because it is easier to focus on computational challenges in OR. With journals like Mathematics of Operations Research that publish theoretical contributions to OR and journals like INFORMS Journal on Computing that publish truly innovative computing research, I am not worried for our field. But I will sit down and try to make my a masterpiece rather than let CPLEX churn on endlessly (Eventually, mbuilding had a power outage, so I never found the optimal solution with CPLEX. I wrote some code that takes just a couple of hours to run. It’s not the most elegant algorithm, but an improvement).