Collaboration in Social Problem-Solving: When Diversity Trumps Network Efficiency
Authors: Diego Noble, Marcelo Prates, Daniel Bossle, Luís Lamb
AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this paper we analyse a recent social problem-solving model and attempt to address its shortcomings. Specifically, we investigate the effects of separating exploitation from exploration in agent behaviors and explore the concept of diversity in such models. We found out that diverse populations outperform homogeneous ones in both efficient and inefficient networks. Finally, we show that agent diversity is more relevant than the strategic behavioral dynamics. This work contributes towards understanding the role of diverse and dynamic behaviors in social problem-solving as well as the advancement of state-of-art social problem-solving models. |
| Researcher Affiliation | Academia | Diego V. Noble, Marcelo O.R. Prates, Daniel S. Bossle and Lu ıs C. Lamb Institute of Informatics, Federal University of Rio Grande do Sul Porto Alegre, 91501-970, Brazil dvnoble@inf.ufrgs.br morprates@inf.ufrgs.br dsbossle@inf.ufrgs.br lamb@inf.ufrgs.br |
| Pseudocode | Yes | Algorithm 1: General algorithm structure used by all models. Algorithm 2: Standard Ring Model procedure. |
| Open Source Code | No | The paper does not provide any statement or link indicating that the source code for the described methodology is publicly available. |
| Open Datasets | No | The paper states: 'The problem instances were generated using a similar process to that of (Mason and Watts 2012).' There is no indication that these generated instances are publicly available or from a standard open dataset. |
| Dataset Splits | No | The paper mentions '1000 independent trials' and describes success criteria, but it does not specify train/validation/test dataset splits in the context of machine learning model training. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., CPU, GPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions the 'Perlin noise method (Perlin 1985)' for generating noise but does not specify any software libraries, packages, or their version numbers used in the implementation of the models or experiments. |
| Experiment Setup | Yes | The problem instances were generated using a similar process to that of (Mason and Watts 2012). The size of the search spaces is fixed at 200 200 integer points i.e. each solution is a pair (x, y) where x and y are integers within the range [0, 200). The coordinates of the signal are chosen randomly and the parameters to generate problem instances are: ω = [2, 3, 4, 5, 6], σ = 9, and persistence parameter ϖ = 0.7. We generated a new instance for each independent trial and used 1000 independent trials. Moreover, our search space is larger and more rugged than the search space employed by Mason & Watts. This leads to a problem that is harder to solve. The peak (best solution) has score of a 100.0. Successful trials are the ones in which any agent achieve a score equal to or higher than 95.0. Moreover, we discarded trials where any agent’s initial score score was higher than or equal to 55.0 in order to avoid the case where an agent is positioned too close to the peak (best solution). We limited the number of iterations to one hundred. We considered the first 32 rounds for this experiment and arranged individuals in a bi-dimensional periodic grid network where all agents have the same degree and are at the same average distance from any other node. By controlling the network, we remove any effect that a network position may have on the agent behavior (Kearns, Suri, and Montfort 2006; Judd, Kearns, and Vorobeychik 2010). We used the population size 64. |