This thesis will evaluate the impact of data-driven methods of eliminating biases learnt by the state-of-the-art neural language models on a case of extractive question answering. The work sets the following objectives: 1. Survey the literature for the known biases that question-answering language models are reported to expose within their predictions. If applicable, the student will extend a list of the known biases for novel ones. 2. Quantify the impact of these biases on the quality of the output of the selected language model. 3. Implement into the training process a selected method mitigating the impact of these biases on model predictions. 4. Evaluate the impact of the implemented bias mitigation method. The experimental setup used for collecting the results will be made publicly available, making the reported results easily reproducible.