Studying Algorithmic Bias in Forensic Source Identification Problems

Presentation Type

Poster

Student

Yes

Track

Forensic Statistics

Abstract

Abstract: This study focuses on forensic source identification, focusing on the hierarchical latent structures that can occur within the relevant source population when using likelihood ratio approaches. We will study systematic algorithmic bias that can occur as measured by Rates of Misleading Evidence in favor of the Prosecutor (RMEP) and the Rate of Misleading Evidence in favor of the Defense (RMED) for each of the subpopulations when the subpopulation structure is not accounted for. This will be done through an extensive simulation study which will identify and characterize subpopulations and quantify forensic evidence. We will be considering varying factors such as the number of subpopulations, mixture proportions, within-source variation, and dimensionality. These parameters add a nuanced layer to the investigation, providing insights into how variations in subpopulation characteristics impact the forensic likelihood ratio. The research illuminates the complex interplay of these factors, enhancing our understanding of forensic evidence assessment and making decisions in hierarchically structured data.

Start Date

2-6-2024 1:00 PM

End Date

2-6-2024 2:00 PM

This document is currently not available here.

Share

COinS
 
Feb 6th, 1:00 PM Feb 6th, 2:00 PM

Studying Algorithmic Bias in Forensic Source Identification Problems

Volstorff A

Abstract: This study focuses on forensic source identification, focusing on the hierarchical latent structures that can occur within the relevant source population when using likelihood ratio approaches. We will study systematic algorithmic bias that can occur as measured by Rates of Misleading Evidence in favor of the Prosecutor (RMEP) and the Rate of Misleading Evidence in favor of the Defense (RMED) for each of the subpopulations when the subpopulation structure is not accounted for. This will be done through an extensive simulation study which will identify and characterize subpopulations and quantify forensic evidence. We will be considering varying factors such as the number of subpopulations, mixture proportions, within-source variation, and dimensionality. These parameters add a nuanced layer to the investigation, providing insights into how variations in subpopulation characteristics impact the forensic likelihood ratio. The research illuminates the complex interplay of these factors, enhancing our understanding of forensic evidence assessment and making decisions in hierarchically structured data.