Introduction¶

Humanitarian tech works only when its risks are surfaced and managed. This outline keeps a hopeful lens while confronting the tensions teams must navigate.

Digital divide and access¶

• Design for low/no connectivity with SMS, USSD, and offline sync; avoid assuming smartphones.
• Budget for device sharing and accessibility: multilingual interfaces, screen-reader support, audio options.
• Co-design with local partners to prevent urban or elite bias.

Privacy and security¶

• Collect minimally; use purpose-limited data with deletion triggers tied to mission end.
• Apply encryption, key rotation, and threat modeling for adversarial environments.
• Offer consent choices that do not penalise beneficiaries.

Bias and fairness¶

• Audit datasets for representation gaps; include edge cases relevant to vulnerable groups.
• Keep human-in/on-the-loop for high-stakes decisions; provide contestation and appeal.
• Track drift and disparate impact over time; sunset models that miss thresholds.

Environmental impact¶

• Prefer efficient models and shared infrastructure; track energy costs of training and inference.
• Use renewable-aligned hosting where feasible; schedule heavy workloads during cleaner grid windows.
• Be transparent about the footprint and trade-offs with mission benefits.

Conclusion¶

Responsible humanitarian tech is deliberate: inclusive design, privacy guardrails, bias monitoring, and climate-aware choices. Naming these risks early keeps the mission - and the people it serves - at the center.