# *Bayesian Modeling of Human-AI Complementarity* This repository contains data **ImageNet-16H** and code to accompany the paper "[Bayesian Modeling of Human-AI Complementarity][1]" published in the Proceedings of the National Academy of Sciences. Please cite this paper when using the data/code. There are three components to this project: - [Behavioral Data][2] - [Machine Classifiers][3] (**NOTE**: there are two types of model prediction files, be sure to read the documentation) - Code - [Bayesian Analysis][4] ## Code There are three JAGS files that are used to run the Bayesian Combination Model: - jags_HH.txt: runs the Bayesian combination model for two human classifiers - jags_HM.txt: runs the Bayesian combination model for one human and one machine classifier - jags_MM.txt: runs the Bayesian combination model for two machine classifiers ## Images The images used for this project are provided in the [Files/Noisy Images][5] folder of this repo. We have provided the noisy images as well as the original images. ### Note About Images The original images come from the [ImageNet Dataset][6]. More specifically, we used images from [ImageNet Large Scale Visual Recognition Challenge 2012 (ILSVRC2012)][7]. **If you would like to use the original images, be sure to read and agree to the [Terms of Access][8] provided by ImageNet.** [1]: https://www.pnas.org/doi/full/10.1073/pnas.2111547119 [2]: https://osf.io/2ntrf/wiki/Behavioral%20Data/ [3]: https://osf.io/2ntrf/wiki/Machine%20Classifiers/ [4]: https://osf.io/2ntrf/wiki/Bayesian%20Analysis/ [5]: https://osf.io/2ntrf/files/osfstorage [6]: https://www.image-net.org/index.php [7]: https://www.image-net.org/challenges/LSVRC/2012/index.php [8]: https://www.image-net.org/download.php