CLIMATE HEALTH AI RESEARCH & DEVELOPMENT

Poverty determines who climate change hits hardest. AI should be built there first.

Climate change is hitting hardest where poverty is deepest and where those communities facing the greatest food and health insecurity are often the least visible to the data systems and AI being developed to respond. That invisibility is both a crisis and an opportunity. AI has matured to reach these communities but only if it’s designed for the conditions they actually face: low connectivity, fragmented data, and rapidly shifting food availability and climate instability.

DHDI’s approach starts from these conditions, not in spite of them. Our current focus is nutrition security for food insecure women and families navigating climate-forced dietary shifts. We integrate community-based research directly into AI development to deliver actionable health guidance and improve health outcomes where they are most at risk.

OUR APPROACH

Two interlocking streams — one designed to feed the other

DHDI’s work operates through two core initiatives. The community engagement stream delivers immediate value — climate-nutrition literacy, food pairing strategies, women’s networks. The AI stream builds on ethically-sourced data generated through that engagement.

STREAM 1

Climate Health Photovox (CHPx)

A 26-session participatory curriculum that translates climate-food-health science into actionable community guidance for food insecure women. Women build climate literacy, smartphone skills, and nutrition knowledge while documenting how forced dietary shifts are reshaping food and health in their communities. Communities gain planning capacity and food-based micronutrient strategies. Images reflecting climate disruptions to local food security are generated through consent-embedded activities.

COMMUNITY RESILIENCE + ETHICAL DATA GENERATION

Climate Health Photovox by DHDI (2024) (CC BY-NC-SA 4.0)

STREAM 2

EpiNu

AI-powered food pairing built on a novel micronutrient database cultivated specifically for food insecure women — not a wrapper on existing composition tables. Designed for community health teams for deployment in low-resource contexts.

COMMUNITY-SCALE NUTRITION AI FOR LOW-RESOURCE CONTEXTS

THE DHDI AI DEVELOPMENT CYCLE

A continuous loop where community benefit and AI improvement reinforce each other — each cycle makes both stronger.

01 ›

CHPx sessions

Women build climate-nutrition literacy and document forced dietary shifts through Climate Health Photovox — a DHDI-developed community engagement strategy

02 ›

Food images with context

Photos of locally consumed foods tagged with local names, preparation state, season, and community — generated through consent-embedded activities

03 ›

AI learns under constraint

EpiNu’s recognition model trains on community-sourced images, operating in low-connectivity, partial-data conditions with structured missingness

04

Micronutrient strategies returned

Food-based pairing guidance for preconception and pregnant women — targeting bioavailability at critical developmental windows with seasonally available foods

CLIMATE HEALTH PHOTOVOX · CHPx

From climate-nutrition science to community-level guidance

CHPx is a 26-session facilitator curriculum developed in consultation with over 150 food insecure women across communities experiencing climate shocks and conflict in South Asia, the Middle East, and East Africa. It bridges three domains that are rarely integrated in a single program: digital literacy for women with low or no literacy, climate science translated into local food security terms, and micronutrient guidance grounded in bioavailability — what the body actually absorbs from the foods available.

The curriculum uses digital storytelling as its method: smartphones become community planning tools. Women document forced dietary shifts, learn food pairing strategies that increase nutrient absorption at near-zero cost, and build cross-community networks for climate alerting and knowledge exchange. Participation is voluntary, with no payment. Currently deployed with civil society partners in eastern DRC.

Pre-Phase

3 sessions

Community Entry

Covers: Mixed-gender community orientation, women’s group formation, consent, baseline assessment, safety and device protocols

Women gain: Community buy-in, shared norms, gender-inclusive framing of nutrition as shared responsibility

Phase 1

11 lessons

Observe · Understand · Document

Covers: Smartphone basics through gallery management; climate impacts on food — floods, heat, crop failure, spoilage; micronutrient science and food pairing; photovoice storytelling arc from observation to narrative

Women gain: Smartphone competence, climate-food-health causal understanding, food pairing strategies for unlocking iron and completing folate-B12 teams, community-authored story set documenting local realities

Phase 2

12 lessons

Connect · Score · Act

Covers: Cross-group networking; EpiNu training for seasonal food scoring and pairing verification; child health under climate stress; early community climate event alerting

Women gain: Cross-community women’s networks, app-confirmed food pairing strategies, climate-optimized nutrition decisions, community early warning capacity, self-sustained network beyond the program

“We don’t ask families to eat different foods or buy things they can’t afford. A squeeze of lemon with beans costs almost nothing but can double the iron a woman’s body absorbs. The curriculum teaches women to see this; the app confirms it with data.”

SCIENTIFIC FOUNDATION

Precision Community Health: Extending precision logic to where poverty structures disease

Standard precision health assumes stable environments, complete data, and individual-level profiling. Under compounding poverty and climate instability, these assumptions fail structurally. Disease variation becomes structured at community-time scale — through biologically sensitive windows where environmental exposures embed durably via epigenetic mechanisms. The communities where this matters most are the least visible to existing systems.

DHDI’s scientific framework uses mechanistic evidence as plausibility guardrails — not rigid specifications. The system learns from partial data through representational and transfer learning approaches, explicitly models structured missingness as informative rather than ignorable, and earns local validity incrementally through governed evidence release. This is not an overlay on existing AI tools. It is a deeper integration between poverty dynamics, climate-driven exposure, and the biological pathways through which both structure chronic disease risk.

White paper forthcoming.

Why existing tools miss this

Food labels ≠ food absorbed

Phytate blocks up to 80% of iron in plant-based staples. Standard guidance reports content, not absorption.

Timing governs consequence

The same nutrients during the first weeks of pregnancy have effects that later supplementation cannot correct.

Climate reshapes diets seasonally

Floods and heat force dietary shifts among the poorest segments — changing what’s available and when.

Invisible ≠ unaffected

Communities generating the highest biological vulnerability are the least visible to formal health and data systems.

ABOUT DHDI

Built to shift who investigates, who benefits, and who decides

The Digital Health Disparities Initiative was founded in 2020 to change how local civil society partners engage with health data — from positions of data collectors to investigators. DHDI operates as a digital community health collab with civil society partners working in health, nutrition and digital literacy domains — building AI infrastructure designed with and for the communities it serves.

Current deployment partners in eastern DRC include women’s health, nutrition, and digital literacy organizations. Formative research draws on consultations with food insecure women across communities experiencing climate shocks and conflict in South and Central Asia, the Middle East, and East Africa.

DHDI Leadership

Sonia Navani, Founder & Director

Implementation scientist with 20 years of expertise building global and local health surveillance systems in fragile settings. Navani is a Senior Fellow at the Blum Center for Developing Economies, UC Berkeley. Field work across 14 countries including Afghanistan, Chad, DR Congo, India, Kenya, Lebanon, Myanmar, Pakistan, South Sudan, Sudan and USA. Previously, she held positions at International Rescue Committee, UNICEF, UNFPA, and Columbia University.

GLOBAL TEAM & OPERATIONS

Oakland / Berkeley Washington DC Colombia DRC Tanzania Australia Pilot Site Team
Legal Status501(c)(3) Nonprofit
Founded2020
HeadquartersOakland, CA