Commentary on artificial intelligence deployment in island public safety contexts and the calibration challenges for Pacific Islander communities.
Commentary on artificial intelligence deployment in island public safety contexts and the calibration challenges for Pacific Islander communities. Free under Creative Commons Attribution 4.0 — no registration required.
Summary of the paper
"When the Algorithm Doesn't Know the Island" is an ISPI commentary on AI-assisted public safety tools in Pacific and island community contexts, including tools used in law enforcement, emergency management and threat assessment.
It argues that many of these tools are developed and tested outside island contexts, and raises the concern that Native Hawaiian, Pacific Islander, Micronesian and Polynesian populations may be underrepresented in the data used to build and validate them. It then explains why small jurisdictions are less able to detect and correct the resulting errors.
What the paper finds
- Validation may not cover island populations. A facial recognition system may perform less accurately on populations it was not validated on; NIST has documented demographic differences in facial recognition accuracy.
- Behavioral scoring can misread culture. An algorithm calibrated on continental organizational norms may flag culturally normal Pacific Island behavior as anomalous.
- Small jurisdictions face an accountability gap. In a large jurisdiction, miscalibration produces enough cases to be statistically visible to oversight bodies. In a small island jurisdiction it affects a few specific people, may never reach detectable volume, and there may be no dedicated oversight body to investigate.
What it recommends
- Require demographic validation documentation, including results for Pacific Islander and Native Hawaiian populations, for AI public safety tools sold to Hawaii and Pacific Island territory agencies.
- Establish mandatory third-party audits of government AI deployment in the Pacific Island territories, with public disclosure of findings.
- Develop small-jurisdiction AI governance frameworks through NIST and CISA that do not assume large-scale accountability infrastructure.
Who it's for
Police and emergency management leaders, procurement officers, legislators and oversight bodies in Hawaii and the Pacific territories, and vendors of public safety AI tools.
This research connects to the 39 Small Island Developing States worldwide through the Samoa Pathway and BPOA+30 framework. Every ISPI research product carries a mandatory SIDS global bridge per the ISPI Research Architecture.
Cite as
Pulley, Warren (2026). When the Algorithm Doesn't Know the Island: AI-Assisted Public Safety Tools in Pacific and Island Community Contexts (ISPI CM-04). Island Security Policy Institute, Honolulu. https://ispiglobal.com/papers-landing/cm04-ai-public-safety https://doi.org/10.5281/zenodo.23449906
BibTeX
@techreport{ispicm042026,
author = {Pulley, Warren},
title = {When the Algorithm Doesn't Know the Island: AI-Assisted Public Safety Tools in Pacific and Island Community Contexts},
institution = {Island Security Policy Institute},
address = {Honolulu, HI},
number = {ISPI CM-04},
year = {2026},
url = {https://ispiglobal.com/papers-landing/cm04-ai-public-safety},
doi = {10.5281/zenodo.23449906}
}Related analysis
AnalysisHPD Chief Lazar Announces Drone Program. The AI Governance Question He Hasn't Answered Yet. →AnalysisBehavioral Threat Assessment in Island Communities — Why Cultural Calibration Changes Everything →AnalysisNative Hawaiian and Pacific Islander Youth in the Justice System — Decades of Disparity, Decades Without Research →PaperCM-03 — Youth ChalleNGe Academy Policy Commentary →PaperCM-05 — Corporate Security and ESG in Island Operations →GuideHawaii Law Enforcement Workforce & Recruitment: Research and FAQ →Pulley, Warren. "AI and Public Safety in Island Communities." Island Security Policy Institute, 2026. https://ispiglobal.com/papers-landing/cm04-ai-public-safety
