This week’s BEACON Researchers at Work blog post is by University of Idaho graduate student Joshua Rubini.
Hi, my name is Joshua Rubini, and I am a graduate (master’s) student in computer science at the University of Idaho. I’m a “non-traditional” student, returning to my education after 12 years in the US Navy, part of which was spent here. I work with Dr. Terence Soule and Dr. Robert Heckendorn on evolved swarm robotics research in the Laboratory for Artificial Intelligence and Robotics.
My interest in artificial intelligence dates back to the summer of 1987, when I played a game called Wizardry, produced by Sir-Tech Software. I eagerly worked my way through the game to the “final” encounter (which was repeatable) and decided that the game just wasn’t smart enough, and that I needed to make a smarter computer game. My father purchased a book on BASIC programming, and at the age of 12 I had a functioning, discreet-room dungeon crawl game, which was the result of some 800 hours and 20000 lines of BASIC code! Needless to say, it was not “smarter” than Wizardry…
Here at the University of Idaho, I took every class I could find relating to AI, but it wasn’t until the Fall of 2008 that I enjoyed my first real exposure to AI research. During our department’s professional seminar series, I listened to Dr. Soule present the research he and Dr. Heckendorn were doing, investigating new methods of evolving teams of AI agents in order to encourage more efficient cooperation solving an exploration problem. I was completely fascinated, and afterward went up and told them how interested I was. The response I received, to my amusement, was, “When can you start?!?” We quickly identified an interesting extension of that experiment that I could work on, and 3 months later it was accepted for publication at GECCO 2009! It was a great experience, and I was hooked. I’ve been working on evolved swarm cooperation and communication ever since.
Earlier this week I was at our department’s professional seminar, which was a talk given by Dr. Carol Taylor, one of my professors years ago. Her talk was about research methodology and the scientific method, and during it she posed the question to the audience, “Is anthropogenic global warming a serious problem?” I looked around, and seeing no one else willing to answer, I raised my hand and said, “I believe it is.” I gave several reasons for my position, after which a couple of other people weighed in, one agreeing with me and one disagreeing. However, both used the word “believe” in their responses.
At this point, the moderator of the seminar series, Dr. Axel Krings, cut in and made an observation. He said,”What is interesting and disturbing about all of your responses is you all used the word ‘believe.’ Belief has no place in a scientific discussion!”
After a minute’s or so reflection on this, I realized that this, in fact, is what my entire semester’s work has really been about. I have been working with Dr. Soule, Travis DeVault, and Melissa Kjelvik on a BEACON-funded project that focuses on educating children about evolution. As a nation, we are sadly lacking in public acceptance of evolution as the driving, central mechanism of biological diversity. Our difficulties here can really be summed up by what Dr. Krings said above; we allow too much “belief” to influence our choices in what we teach students about biology. The other major hurdle is that evolution, as a process, is difficult to “see” working.
Our program attempts to tackle the second problem directly, with a fun, engaging application that allows student to not only see evolution taking place in front of them, but also allows them to “tinker” with the mechanisms and forces that drive evolution. Students can see how these processes work together to allow populations to evolve and adapt to changing conditions.
The program is called “The Ladybug Game” and shows a simulated leaf with a ladybug and a bunch of aphids. The ladybug runs around the leaf, eating aphids that it sees in front of it. Each time an aphid is eaten, another is spawned, inheriting traits from its parent and mutating those traits slightly. The inherited traits are color and strategy, where the “strategy” of each aphid is defined by a simple neural network which decides how much to change its heading and speed. The chance that an aphid is “seen” by the ladybug is dependent on how closely that aphid matches the background leaf color. In this way, the ladybug acts as the “selection” function, eating aphids that haven’t blended in well enough into the background.
The program focuses on three primary facets of evolution: selection, inheritance, and variation. In three of the lessons, students get to watch two separate populations, one with all of the evolutionary machinery running, and one with a single mechanism removed. For example, in lesson 3, one of the populations has variation removed entirely, such that new aphids are exact copies of their parent. It quickly becomes apparent that these aphids cannot adapt at all if the background color of the leaves is changed, since there is no way to introduce any sort of random change once the population converges on a single color. Similar to this lesson, two more lessons each show what happens if you remove color selection or inheritance.
Working on this program has been a very rewarding experience for me, as I feel that its use could make a very real impact on how the next generation of investigators understands the fundamentals of evolution. I have very much enjoyed working with Terry and Melissa on this, and look forward to seeing it in classrooms in the future.
Current version of the program can be found at:
www2.cs.uidaho.edu/~rubi4714/LBapp/index.html
A detailed description of each lesson is at:
www2.cs.uidaho.edu/~rubi4714/Teaching.html#Ladybug
For more information about Josh’s work, you can contact him at rubi4714 at vandals dot uidaho dot edu.
