@{, author = {Remi van Trijp, Ines Blin},booktitle = {Foundations for Meaning and Understanding in Human-Centric AI},editor = {Steels, Luc},publisher = {Venice International University},title = {Narratives in Historical Sciences},year = {2022},url = {https://doi.org/10.5281/zenodo.6666820},abstract = {

Historical sciences such as geology, evolutionary biology or history try to o↵er causal explanations for non-recurrent phenomena (e.g. how the Grand Canyon was formed, how the human eye evolved, or what caused the Second World War), typically using incomplete and fragmentary evidence from the past. Even though these sciences make use of general frameworks such as the Theory of Evolution, they have to work out the specifics of each case and hence they cannot simply apply general laws and deductive reasoning. Instead, causal explanations are expressed in narrative form. While such explanations have long been considered to be “less scientific”, there is now a growing aware- ness that narrative explanations go beyond mere description and also have the potential for empirical testing. This Chapter explores how narratives are used by historical scientists, and how human-centric AI systems may assist scien- tists in constructing more precise and testable narratives so they can achieve a deeper understanding of society. It presents a first prototype that takes as its case study the French Revolution (1789–1799).