Responsible AI
Ethics & Responsible Use
“Privacy-safe, human-in-the-loop decision support for dengue preparedness.”
DengueOps AI is designed as a public health preparedness prototype, not a diagnostic or autonomous decision-making tool. It uses aggregated and synthetic demonstration data to show how outbreak forecasts can be translated into readiness alerts and action priorities without processing patient-level information.
Design intent: This prototype demonstrates a privacy-safe, transparent, and human-in-the-loop approach to AI-enabled public health preparedness. It prioritises responsible simulation over unsupported claims of live deployment.
Design Principles
Ethical Design Principles
Six core ethical commitments are embedded in the prototype design.
No Patient-Level Data
The prototype does not collect, store, or process identifiable patient-level records. Forecasting and operational outputs are based on aggregated/demo data structures.
Privacy-Safe Demonstration
Facility readiness, inventory, and dengue-specific bed values are synthetic demonstration values. This prevents unauthorised use of sensitive hospital operations data.
Human-in-the-Loop Decision-Making
Outputs are advisory. Final decisions remain with public health officials, hospital administrators, and qualified authorities.
No Clinical Diagnosis
The system does not diagnose dengue, recommend individual treatment, or replace clinicians.
Transparent Uncertainty
Forecast uncertainty is shown using best, expected, and worst-case scenarios so decision-makers can plan cautiously.
Bias and Vulnerability Safeguards
The vulnerability-gated priority score prevents static vulnerability factors from permanently dominating response priorities when forecasted dengue risk is low.
Data Ethics
Data Ethics and Data Boundaries
Clear boundaries define what is real, what is anchored from public sources, and what is entirely synthetic demonstration data.
| Data Element | Status | Note |
|---|---|---|
| Dengue case data | Synthetic / Demo | Demo aggregate values for pipeline demonstration. |
| Climate data | Synthetic / Demo | Synthetic/demo or public-style aggregate values. |
| Public hospital names | Public Anchor | Used as geographic facility anchors where available. |
| General bed capacity | Public Anchor | Public reference anchor where available; synthetic otherwise. |
| Dengue-specific beds | Synthetic / Demo | Synthetic demonstration values only. |
| Current dengue occupancy | Synthetic / Demo | Synthetic demonstration values only. |
| NS1/RDT stock | Synthetic / Demo | Synthetic demonstration values only. |
| IV fluid stock | Synthetic / Demo | Synthetic demonstration values only. |
| Baseline daily consumption | Synthetic / Demo | Synthetic demonstration values only. |
| Patient-level records | Not Used | Not collected, processed, or stored. |
Exact statement: Facility names and general bed-capacity anchors are based on public/government references where available. Dengue-specific bed allocation, current occupancy, NS1/RDT stock, IV fluid stock, and consumption values are synthetic demonstration values.
Safety Boundaries
What the System Does Not Do
Clear negative constraints are part of the responsible design.
It does not — Diagnose dengue.
It does not — Recommend individual treatment.
It does not — Replace clinical judgment.
It does not — Claim access to live hospital stock or bed data.
It does not — Automatically trigger public health action.
It does not — Provide official public warnings.
It does not — Process patient-identifiable data.
Output Translation
Responsible Output Design
Technical model outputs are translated into operational preparedness metrics before reaching decision-makers.
Technical layer
- · MAE / RMSE
- · Forecast uncertainty
- · Model comparison
- · Feature lags
Operational layer
- · Risk level
- · SDH
- · Bed gap
- · Priority score
- · Recommendations
Decision layer
- · Human review
- · Facility planning
- · Vector-control prioritisation
- · Contingency planning
RMSE and model metrics are included for technical validation and evaluator transparency; operational users receive translated preparedness outputs rather than raw model diagnostics.
User Roles
User Roles and Ethical Responsibilities
Each user role interacts with the prototype in a defined, bounded capacity.
MIS / Data Officer
- · Maintains the analytics data pipeline
- · Validates input files before scheduled runs
- · Runs scheduled pipeline updates
Public Health Analyst
- · Reviews forecast behaviour and uncertainty ranges
- · Checks assumptions and data limitations
- · Interprets outbreak patterns cautiously
Hospital Administrator
- · Uses readiness alerts for facility planning
- · Confirms real stock and bed status before taking action
- · Does not treat synthetic values as definitive
City Corporation / Vector-Control Team
- · Uses zone priority score as one input among many
- · Confirms local field conditions before deployment
- · Does not act on prototype output alone
Technical Evaluator
- · Reviews model validation, limitations, and assumptions
- · Assesses pipeline design and decision-support logic
- · Does not evaluate as a live clinical system
Future Deployment
Future Deployment Ethics
Real operational deployment of this system would require formal governance and institutional safeguards beyond the prototype scope.
Institutional approval and governance
Official data-sharing agreements
Privacy review and data protection assessment
Facility-level validation of readiness logic
Access control and role-based permissions
Audit logs for all data and forecast actions
Regular model monitoring and recalibration
Public health governance and accountability chain
Clear legal accountability framework
DengueOps AI demonstrates a privacy-safe, transparent, and human-in-the-loop approach to AI-enabled public health preparedness. The prototype prioritises responsible simulation over unsupported claims of live deployment.
Ethics Summary — DengueOps AI · IEEE ICADHI 2025
DengueOps AI — Simulation-Based Dengue Surge Preparedness Decision Support for Dhaka South. Prototype only. Not for clinical or official public health use.