Friday, March 27, 2015

Bumble Bee Colonies

When reading papers about agent-based models / individual-based models, I am always curious if the model can be reproduced from the description in said paper. Specifically whether there is sufficient information in the paper to reproduce the model and the results. Often we task students with such a task as a learning exercise and I thought it would be nice to show you one such example done by Dale Brearcliffe.

The model that was reproduced is entitled "The Ontogeny of the Interaction Structure in Bumble Bee Colonies: A MIRROR Model" by Hogeweg and Hesper (1983) which explored whether or not an individual-oriented model of population dynamics and simple bumble bee behaviors could produce the ontogeny of the social interaction of the colony? The original model used MIRSYS, a program written   in INTERLISP. Dale was able to take the paper and reproduce the model using NetLogo, but not replicate the results (click here to read more and download the model).

Graphical User Interface of Reproduced  Model

Nest composition and development in the original model (Source: Hogeweg and Hesper, 1983).
Nest composition and development in the reproduced model.

Full Reference:
Hogeweg, P. and Hesper, B. (1983), 'The Ontogeny of the Interaction Structure in Bumble Bee Colonies: A MIRROR Model', Behavioral Ecology and Sociobiology, 12(4): 271-283. 

For those interested in reproduction and replication have a look at the following articles:
Axtell, R., Axelrod, R., Epstein, J.M. and Cohen, D. (1996), 'Aligning Simulation Models: A Case Study and Results', Computational and Mathematical Organization Theory, 1(2): 123-141. 
Drummond, C. (2009), 'Replicability is Not Reproducibility: Nor is it Good Science', Proceedings of the 4th Workshop on Evaluation Methods for Machine Learning at the 26th International Conference on Machine Learning, Montreal, Canada.
Wilensky, U. and Rand, W. (2007), 'Making Models Match: Replicating an Agent-Based Model', Journal of Artificial Societies and Social Simulation, 10(4): 2, Available at


Ellen said...

Wow, nice blast from the past: I was another one of your students who reimplemented the Hogeweg and Hesper model (in Repast) a few years back, but kudos to Dale for producing a version that looks to be far more intuitive and user-friendly!

It's very interesting that neither of us (out of my sample set of two) was able to replicate the original results. Thanks for the article links.

Andrew Crooks said...

Hi, Never a easy task.