the CMU mocap dataset is a collection of 2537 actions like walking, talking, jumping etc. generated from motion capture. it is used, among others, in machine learning settings, and can there become a source of (well-hidden) bias eg. if all those motions were only generated from male, or healthy, or typically western people. the goal of this master thesis is to discuss how bias is defined in the context of such a dataset, how the dataset can be evaluated for various biases, and conduct exemplary evaluations for one or more bias dimensions. data set bias evaluation is an exciting new research area that is steadily gaining in importance.