EAA2020: Abstract
Abstract is part of session #464:
Title & Content
Title:
Patterns of Trauma -
Using AI to distinguish interpersonal violence from accidental injury
Using AI to distinguish interpersonal violence from accidental injury
Content:
Distinguishing interpersonal violence from accidental injuries is a vital research component to understand the emergence, complexity and variety of fracture patterns in a socio-bioarcheological and modern context. It offers not only insights to individual personal lives but also the presence, occurence and sanctioning of violence within a society.
Whilst these patterns can be observed in the past and present their expression changes throughout culture and time. Not only is the expression of violence culture and time specific, lifestyles and modes of subsistence have changed, too.
Therefore, they need to be examined carefully - as differences in the mode and intend of force, be it interpersonal or accidental - can be distinguished by specialists. However, besides sharp and projectile trauma, singular blunt force fractures are often hard to interpret. In general, "pathognomonic" fractures of small scale, non war related, interpersonal violence are extremely rare and their interpretation will and should always involve expert opinions. The potential of ML allows to use a tagged dataset for training and generate models based on the location of the fractures, their lethality, biological sex, age, and burial place of the individuals. It increases in sensitivity through interations predicts more accurately with time.
This talks presents a proof of concept to use AI based models on published archaeological and modern cases to replicate the divide between accidental injuries and lethal interpersonal violence. We hope to explore the potential of these methods with a critical view on their mechanics and how they can aid to explore new hypotheses, especially considering the vast record of bioarchaeological human remains for population studies. The project aims to expand its transdisciplinary research to the distinction of patterns of interpersonal violence in the future.
Whilst these patterns can be observed in the past and present their expression changes throughout culture and time. Not only is the expression of violence culture and time specific, lifestyles and modes of subsistence have changed, too.
Therefore, they need to be examined carefully - as differences in the mode and intend of force, be it interpersonal or accidental - can be distinguished by specialists. However, besides sharp and projectile trauma, singular blunt force fractures are often hard to interpret. In general, "pathognomonic" fractures of small scale, non war related, interpersonal violence are extremely rare and their interpretation will and should always involve expert opinions. The potential of ML allows to use a tagged dataset for training and generate models based on the location of the fractures, their lethality, biological sex, age, and burial place of the individuals. It increases in sensitivity through interations predicts more accurately with time.
This talks presents a proof of concept to use AI based models on published archaeological and modern cases to replicate the divide between accidental injuries and lethal interpersonal violence. We hope to explore the potential of these methods with a critical view on their mechanics and how they can aid to explore new hypotheses, especially considering the vast record of bioarchaeological human remains for population studies. The project aims to expand its transdisciplinary research to the distinction of patterns of interpersonal violence in the future.
Keywords:
Machine Learning, Trauma Patterns, Concept
Downloads:
authors
Main authors:
Chiara G. M. Girotto1
Co-author:
Henry C. W. Price2
Martin Trautmann3
Martin Trautmann3
Affiliations:
1 Independent Researcher
2 Imperial College London
3 A&O Praxis für Bioarchäologie, München
2 Imperial College London
3 A&O Praxis für Bioarchäologie, München