On Wednesday November 18 we had another exciting talk in our HeiCADLecture series. Dr Bennett Kleinberg from Tilburg University and University College London discussed both promises and challenges in data science using the practical example of crime research.
By means of his current study on the temporal evolution of a far-right internet forum, he illustrated how well phenomena such as the frequency of extremist posts by particularly active users can be modelled mathematically. One promise in the field of data science is to be able to research human behaviour on a large scale and in high granularity due to the abundance of data, tools and programming expertise.
On the other hand, Dr Kleinberg pointed out methodological limits. Using two studies on label validity for emotions that were derived from short texts and tweets, he demonstrated that these often do not match the true underlying emotions. Therefore, their use is problematic. Overall, Dr Kleinberg pleads for not losing the focus on fundamental methods in data science in order to create comparability, unambiguousness and transparency.