Research Scope
Subject Classification
Fifteen Caribbean subject areas and twelve machine-learning method tags. Every submission to the journal, the preprint server and the student section carries one primary code and up to three cross-lists.
Classification
How a Submission Is Classified
Classification lets a reader follow one slice of a field without reading all of it. arXiv has used subject codes like cs.LG and stat.ML since 1991: a primary archive that decides where a paper is announced, plus cross-lists that surface it to neighbouring communities. The scheme below borrows that machinery and applies it to a region instead of a discipline.
There are two axes, because a single-axis scheme breaks the moment a paper is about hurricanes and about time-series forecasting. The subject axis records what the work is about in the Caribbean. The method axis records what machine learning it uses. Everything published by AIC carries a code on both.
- One primary subject code, beginning
cai., which decides which section editor handles the submission and where it is listed. - One primary method tag, beginning
ml., which decides which reviewers are approached. - Up to three cross-list codes from either axis, which surface the work to adjacent readers without changing who owns the decision.
Authors propose their own codes at submission. Editors adjust them where a paper is filed in the wrong place, and tell the author when they do. Misclassification is never a ground for rejection.
Two conditions apply to every code on this page. The work must apply machine learning or artificial intelligence rather than only discuss it, and it must engage the Caribbean, its diaspora, or the small-state and Small Island Developing State conditions the region shares with the Pacific, the Indian Ocean and comparable economies. A technically strong paper that meets neither condition is out of scope, and desk review will say so within five business days rather than send it to reviewers.
Axis One
Subject Areas
Fifteen areas covering the domains where AI research in this region can change a decision. Each carries a scope note and the kind of work we expect to see filed under it.
| Code | Subject area | Scope |
|---|---|---|
cai.CLM |
Climate, Weather and Disaster Risk | Hurricane track and intensity prediction, storm surge and flood mapping, landslide risk, drought and water stress forecasting, coral and reef monitoring, sea-level and coastal change, early warning and evacuation systems, post-event damage assessment, parametric trigger design, climate-adjusted infrastructure planning. |
cai.FIN |
Finance, Insurance and Risk | Credit scoring for thin-file and unbanked borrowers, alternative data underwriting, catastrophe and actuarial modelling, parametric and index insurance, remittance flows, fraud and anti-money-laundering detection, sovereign debt and fiscal risk, capital markets and pensions in shallow markets, financial inclusion. |
cai.GOV |
Public Policy, Governance and Law | AI inside public administration, revenue and customs risk, benefits and service triage, court and case-law analytics, legislative modelling, procurement integrity, evaluation of deployed government systems, regulatory design and algorithmic accountability for small states. |
cai.SME |
Small Business, Trade and Informal Economies | Demand and inventory forecasting for micro-enterprise, market access, informal sector measurement, supply chain and logistics across archipelagos, digital payments, AI adoption studies among small firms, trade and customs analytics. |
cai.INF |
Infrastructure, Compute and Data Systems | Regional compute and cloud strategy, sovereign and national language models, data governance and interoperability, national data infrastructure, connectivity, model serving under bandwidth, power and cost constraints, MLOps for small institutions. |
cai.MED |
Health, Medicine and Life Sciences | Diagnostic imaging validated on regional populations, non-communicable disease risk models, outbreak and vector-borne disease surveillance, health system triage and capacity planning, genomics and structural biology, natural products and Caribbean biodiversity. |
cai.AGR |
Agriculture, Fisheries and Food Security | Yield, pest and disease prediction, precision irrigation under water scarcity, soil and land use, fisheries stock assessment and illegal fishing detection, sargassum forecasting, food import dependence and price transmission. |
cai.TUR |
Tourism, Mobility and the Blue Economy | Arrival and demand forecasting, dynamic pricing and revenue management, cruise and port logistics, visitor sentiment analysis, transport and mobility modelling, marine spatial planning, coastal carrying capacity. |
cai.CUL |
Language, Culture and Creative Industries | Creole and heritage language corpora, speech recognition and translation for regional languages, dialect and code-switching modelling, archive digitization and search, music, film and design economics, generative media and cultural heritage. |
cai.EDU |
Education, Skills and the Workforce | Learning analytics on regional examination data, adaptive tutoring in low-resource classrooms, dropout and attainment prediction, labour market and skills forecasting, automation exposure of Caribbean occupations, brain drain, AI literacy and curriculum research. |
cai.NRG |
Energy, Water and Utilities | Grid stability under high renewable penetration on isolated grids, load and outage forecasting, restoration sequencing after storms, non-revenue water detection, microgrid and storage optimization, demand response, energy poverty measurement. |
cai.SEC |
Safety, Security and Border Systems | Crime and violence analytics and its limits, road traffic safety, maritime domain awareness, search and rescue, border and immigration systems and their error rates, disaster response coordination, cyber defence for small institutions. |
cai.MIG |
Migration, Diaspora and Demography | Migration flow modelling and drivers, diaspora investment and network analysis, remittance behaviour, population projection for ageing and shrinking populations, census and administrative data linkage under privacy constraints. |
cai.SOC |
Ethics, Society and Human Rights | Measured bias in models trained without the region, surveillance and privacy, consent and data sovereignty, labour effects of automation, participatory research design, AI and reparative justice, environmental cost of AI in vulnerable states. |
cai.MTH |
Methods, Benchmarks and Open Datasets | New Caribbean datasets and benchmarks, evaluation protocols for low-data regimes, transfer and domain adaptation studies, reproducibility and replication work, measurement and survey methodology, tooling that other regional researchers can reuse. |
Axis Two
Method Tags
Twelve tags describing the machine learning a paper uses. The tag decides which reviewers are approached, so choose the one a specialist in the method would recognize.
| Code | Method | Covers |
|---|---|---|
ml.SUP | Supervised and Statistical Learning | Regression, classification, gradient boosting, tree ensembles, calibration, classical statistical modelling applied to prediction |
ml.DEE | Deep Learning and Architectures | Neural network design, transformers, training methods, fine-tuning, distillation, quantization, efficiency work |
ml.NLP | Language, Speech and Creole NLP | Text and speech processing, low-resource and creole language technology, translation, information extraction, corpora construction |
ml.VIS | Computer Vision | Image and video understanding, detection and segmentation, document and archive imaging, medical imaging, drone and aerial imagery |
ml.GEN | Generative Models | Large language models, diffusion and image generation, synthetic data, retrieval augmented generation, evaluation of generative output |
ml.AGT | Agentic and Multi-Agent Systems | Tool-using agents, planning, multi-agent simulation, agent-based modelling of social and economic systems |
ml.RLX | Reinforcement Learning and Control | Policy learning, sequential decision making, operations research crossovers, control of physical and logistical systems |
ml.CAU | Causal Inference and Econometrics | Causal identification, quasi-experimental design, policy evaluation, structural and econometric modelling with learned components |
ml.TSA | Time Series and Forecasting | Univariate and multivariate forecasting, anomaly detection, nowcasting, seasonality and regime change, uncertainty quantification |
ml.GEO | Geospatial and Earth Observation | Satellite and remote sensing pipelines, spatial statistics, GIS integration, land cover and change detection, spatial interpolation |
ml.FED | Federated, Edge and Low-Resource ML | Training and inference under constrained compute, bandwidth or power, on-device models, privacy-preserving and federated learning |
ml.XAI | Interpretability, Fairness and Evaluation | Explanation methods, bias measurement and mitigation, robustness testing, benchmark design, audit methodology, model documentation |
Examples
Choosing Your Codes
In each example the primary subject code is the domain the finding belongs to, and the primary method tag is whatever a reviewer needs to be able to check.
Forecasting outage duration after a hurricane
Primary: cai.NRG and ml.TSA
Cross-list: cai.CLM, ml.GEO
The finding is about the electricity grid, so it files under energy rather than climate, even though the trigger is a storm. Climate readers still find it through the cross-list.
A speech dataset and recognizer for Jamaican Patois
Primary: cai.CUL and ml.NLP
Cross-list: cai.MTH, cai.SOC
A dataset paper cross-lists to methods so that other builders find it, and to ethics because language data raises consent and ownership questions that reviewers should weigh.
Testing a credit model built abroad on local applicants
Primary: cai.FIN and ml.XAI
Cross-list: cai.MTH, ml.SUP
The contribution is the fairness and transfer audit, not the model, so the method tag is interpretability and evaluation rather than supervised learning.
Revising the Scheme
A classification scheme that never changes stops describing the field it was built for. arXiv has added, split and retired categories repeatedly across three decades, in response to what researchers were submitting.
The AIC scheme is reviewed once a year by the Editor-in-Chief and the section editors, against the distribution of submissions received. A code that attracts almost nothing is merged. A cross-list that repeatedly carries more traffic than its parent becomes a subject area of its own. Changes are announced before an edition opens, never during one, and existing published records keep the codes they were assigned.
If your work fits nowhere on this page, write to editor@caribbeanaijournal.org with the paper and a one-line case for the code you think is missing. Proposals are considered at the annual review, and sooner where a whole line of regional research is going unclassified.
Inaugural Edition
Submit to the Inaugural Edition
The first edition accepts submissions across all fifteen subject areas, in three tracks, with no article processing charge. Submissions close 14 September 2026.