Mapping the uncertainty of 19th century West African slave origins using a Markov decision process model
share
The advent of modern computers has added an increased emphasis on channeling computational power and statistical methods into digital humanities. Including increased statistical rigor in history poses unique challenges due to the inherent uncertainties of word-of-mouth and poorly recorded data. African genealogies form an important such example, both in terms of individual ancestries and broader historical context in the absence of written records. Our project aims to bridge the lack of accurate maps of Africa during the trans-Atlantic slave trade with the personalized question of where within Africa an individual slave may have hailed. We approach this question with a two part mathematical model informed by two primary sets of data. We begin with a conflict intensity surface which can generate capture locations of theoretical slaves, and accompany this with a Markov decision process which models the transport of these slaves through existing cities to the coastal areas. Ultimately, we can use this two-step approach of providing capture locations to a historical trade network in a simulative fashion to generate and visualize the conditional probability of a slave coming from a certain spatial region given they were sold at a certain port. This is a data-driven visual answer to the research question of where the slaves departing these ports originated.
READ FULL TEXT