Working paper
Inside the folder
Author
Authored working paper · 2026
AI governanceDigital trustAutomated exclusionInclusive security
Description
Examining how AI security and governance can protect a system while making it harder for people in low-resource environments to access services or recover from an automated decision.
Working paper
The research question
When AI makes a service more secure, who pays the cost of that security, and can the people affected understand, challenge, and recover from a wrong decision?
My contribution
- Examined identity, fraud detection, authentication, and digital-service systems through the lens of automated exclusion
- Introduced AI Security Burden as a way to account for the technical, cognitive, economic, and procedural costs placed on users
- Proposed a Last-Mile AI Governance Framework for evaluating security alongside access, redress, and recovery
Outcomes & evidence
- A framework organized around proportional security, inclusive authentication, explainable decisions, meaningful human redress, and equitable recovery
- An argument for assessing who carries the cost of an incorrect AI decision, not only model accuracy or system-level security
- A public working paper available on Zenodo
Approach & methods
- Compared AI-enabled identity, authentication, fraud detection, and digital-service systems, paying attention to low-connectivity and low-resource contexts.
- Examined assumptions such as reliable connectivity, individual device ownership, formal identification, digital literacy, and abundant user data.
- Developed a conceptual governance framework that treats access, explanation, human redress, and recovery as part of security design.
Scope & limitations
- This is a working-paper framework. The library does not present it as an accepted conference paper or a quantitatively validated intervention.
- The summary describes the public abstract; it does not add study populations, effect sizes, or evaluation results that the abstract does not report.