You Get What You Share: Incentives for a Sharing Economy
Authors: Sreenivas Gollapudi, Kostas Kollias, Debmalya Panigrahi2004-2011
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we perform simulations which show that the SHAPLEY method improves on EQUAL in realistic settings, with respect to three metrics that measure our desiderata for forming sharing groups: performance (measured by the social welfare of the partition), participation (measured by the number of sharing groups of size k), and fairness (measured by the quality of the weakest group). |
| Researcher Affiliation | Collaboration | Sreenivas Gollapudi Google Research Kostas Kollias Google Research Debmalya Panigrahi Duke University |
| Pseudocode | No | The paper does not contain any explicit pseudocode or algorithm blocks. |
| Open Source Code | No | The paper mentions a "full version" with a PDF link, but does not provide any link to open-source code for the methodology or state that code is available. |
| Open Datasets | No | In our setup, we begin with |N| = 5,000 agents and a hypothetical set of |P| = 20 resource types that all agents can utilize in groups of size at most k = 5. |
| Dataset Splits | No | The paper describes a simulation setup rather than a traditional machine learning experiment with dataset splits (training, validation, test sets) for model reproduction. |
| Hardware Specification | No | The paper describes the simulation setup but does not provide any details regarding the hardware specifications used for running the experiments. |
| Software Dependencies | No | The paper does not list any specific software dependencies with version numbers. |
| Experiment Setup | Yes | In our setup, we begin with |N| = 5,000 agents and a hypothetical set of |P| = 20 resource types that all agents can utilize in groups of size at most k = 5. In each simulation, we begin with each agent in a sharing group by herself. Then, in each round we let the agents deviate (one by one) to a sharing group that they prefer until no such deviations are possible, i.e., until an NE is reached. We use three metrics to evaluate the quality of the partition reached by the agents. The first one is the natural performance metric, which measures the social welfare of the resulting partition. The second one is a participation metric, which counts the number of sharing groups for which the cardinality constraint is tight. As we will see, SHAPLEY almost always yields partitions such that all groups are full, which also validates our study of the BNE as an important equilibrium concept in this setting. Finally, we focus on a fairness metric, which measures the minimum social welfare extracted by an agent, i.e., the number of resources available in the weakest group. In the following sets of experiments we use skewed distributions that follow power laws. |