A normative theory of social conflict
Authors: Sergey Shuvaev, Evgeny Amelchenko, Dmitry Smagin, Natalia Kudryavtseva, Grigori Enikolopov, Alex Koulakov
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | To understand its underlying principles, we collected behavioral and whole-brain neural data from mice advancing through stages of social conflict. We modeled the animals interactions as a normal-form game using Bayesian inference to account for the partial observability of animals strengths. We find that our behavioral and neural data are consistent with the first-level Theory of Mind (1-To M) model where mice form primary beliefs about the strengths of all mice involved and secondary beliefs that estimate the beliefs of their opponents. Our model identifies the brain regions that carry the information about these beliefs and offers a framework for studies of social behaviors in partially observable settings. |
| Researcher Affiliation | Academia | 1Cold Spring Harbor Laboratory, Cold Spring Harbor, NY, USA; 2Department of Anesthesiology and Center for Developmental Genetics, Stony Brook University, Stony Brook, NY, USA; 3Institute of Cytology and Genetics, Siberian Branch of Russian Academy of Sciences, Novosibirsk, Russia. |
| Pseudocode | Yes | Algorithm 1 Game theory optimal actions; Algorithm 2 Bayesian belief initialization; Algorithm 3 Bayesian belief update; Algorithm 4 Model parameter inference; Algorithm 5 Belief correlates in the brain. |
| Open Source Code | No | The paper does not contain any statement or link indicating that the source code for the described methodology is publicly available. |
| Open Datasets | No | The paper describes the collection of 'large-scale data on behavior and wholebrain neural activity' and mentions 'training data (88 mice participating for 22 days)', but it does not provide concrete access information (e.g., URL, DOI, repository) for this dataset. |
| Dataset Splits | No | The paper mentions 'training data' and 'testing data' but does not explicitly state the use of a separate 'validation' dataset split or cross-validation for hyperparameter tuning. |
| Hardware Specification | Yes | All computations were performed with Matlab R2022b. The mouse data fits were performed with a 3 GHz Intel Core i7-based Dell laptop computer. The simulated data fits were performed with a 3 GHz Intel Xeon-based Supermicro computer server in 40 parallel threads. |
| Software Dependencies | Yes | All computations were performed with Matlab R2022b. |
| Experiment Setup | Yes | To fit the model parameters, we minimized the NLL regularized with the l2 norm of the model arguments. We chose the regularization coefficient such that it resulted in the best fits in a simulated experiment. (...) We observed a local minimum of the NLL at the parameter values σ1 = 3 1g, σ2 = 10 1g, βo = 6 1, βa = 3 1, α = 3 1, A = 10 1, and ε = 0.8 0.2. |