Special Call: AI Methods for Population Research

Overview

The Hopkins Population Center (HPC) Scientific Core invites HPC Faculty Associates to apply to a special Method Improvement Fund (MIF) call focused on developing or adapting artificial intelligence (AI) methods for population research. This special call seeks short, high-impact projects that advance methodological capabilities for working with complex and novel data types (e.g., text, audio, images, video, administrative, and high-frequency digital traces), improve measurement and indicators, or create analytic strategies that meaningfully advance one or more of HPC’s three primary research areas:

• Poverty and Inequality
• Sexual and Reproductive Health
• Family, Maternal and Child Health

The MIF is intended to support method development or adaptation that will lead to an external grant proposal or a scientific paper and produces reusable tools, code, benchmarks, or documented workflows that can benefit other HPC Associates. For this particular call, awards are dependent on the disbursement of the anticipated NICHD P2C Bridge Funds.

Key dates and award information:

• Application deadline: July 15, 2026; decisions are expected to be announced around August 15, 2026
• Project period: 6 months from the award date (earliest start date: September 1, 2026)
• Number and amount: 2 awards of $6,000 each

Scope and Priorities

Eligible projects should use AI (including but not limited to machine learning, deep learning, natural language processing, computer vision, multimodal models, representation learning, causal machine learning, federated learning, or foundation model adaptation) to address methodological challenges that are directly relevant to at least one HPC primary research area listed above. Applicants are encouraged to form multidisciplinary teams and to plan for delivering reproducible code and clear documentation that can be shared with HPC Associates subject to data use limitations.

Priority will be given to:

• Projects that clearly link the AI method to improved measurement, construct validity, or analytic strategy for population research
• Multidisciplinary teams (e.g., population scientists partnering with data scientists/computer scientists)
• Projects that will produce open-source code, reproducible pipelines, documentation, or demonstrations (subject to data use restrictions)
• Early-career investigators and proposals that provide training opportunities for graduate or postdoctoral trainees
• Feasible, well-scoped projects that can be completed within 6 months and with the requested budget

Examples of appropriate AI-focused MIF projects:

• Developing and validating natural language pipelines to extract reproductive health indicators from clinical notes or social media while evaluating bias and validity across subpopulations (relevant to sexual & reproductive health or maternal and child health)
• Training and evaluating computer vision models to detect household-level indicators of living standards from aerial or street imagery for poverty and inequality measurement, including calibration/robustness checks (relevant to poverty and inequality)
• Adapting foundation models for small-sample causal inference or heterogeneity analysis in longitudinal cohort data related to maternal and child outcomes (relevant to family, maternal and child health)
• Building multimodal representations combining survey, text, and sensor data to improve measurement of parental caregiving or child developmental stimulation (relevant to family, maternal and child health)
• Developing privacy-preserving/federated learning workflows to train models across administrative datasets from multiple jurisdictions while preserving confidentiality (relevant methods development potentially deployable to all 3 primary areas of interest)
• Creating synthetic data generation, e.g. digital twins, and validation pipelines to support method testing for rare reproductive health outcomes
  

Eligibility

• All HPC Faculty Associates are eligible
• Projects must align with at least one HPC primary research area mentioned above
• Intervention or clinical trials are out of scope; method development that supports population research is the focus.

Proposal format and content

Proposals should be submitted as a single PDF and must be no more than two single-spaced pages (11-point Arial, half-inch margins), plus NIH biosketch(es) and IRB documentation, if applicable, in the appendix. Include the following specific sections:

1. Project title and 1–2 line subtitle describing the AI method component
2. Research question and objectives (1 paragraph)
3. Relevance to HPC priorities: specify which primary research area(s) the project addresses and how the method will improve measurement or analysis in that domain or across domains
4. Current methods and rationale: brief description of existing/traditional approaches and why AI methods are needed or advantageous
5. Proposed AI methods and outputs: describe the AI approach, model types, data inputs (format and access), expected deliverables (e.g., code repository, trained models, documentation, benchmarks, synthetic datasets), and evaluation plan (including assessments of validity, bias, fairness, generalizability, and reproducibility where relevant)
6. Team and roles: list PI, co-PI(s), and other key personnel, noting any data science/computer science collaborators and trainees to be supported
7. Timeline (6-month) with milestones specified
8. Budget and brief justification (not counted in the two pages). Use broad line items (e.g., graduate RA, programmer, data access fees, computing/cloud credits)
9. Appendix: current NIH biosketch for PI and key personnel; IRB approval if human-subject data are involved

Allowable and non-allowable costs

Allowable:

• Graduate research assistant support
• Programmer/data scientist time
• Fees for data access or data use agreements
• Computing costs or cloud credits necessary for model training (please justify; HPC may request documentation)
• Subsidies for method conference presentation to disseminate methods

Not allowable:

• Fees for consultant external to Hopkins
• Publication fees
• Computer hardware purchases
• Travel to professional meetings (general travel not allowed unless specifically for a methods dissemination workshop)

Review process and criteria

• Eligible proposals will be reviewed by 2 HPC Associates with relevant domain and/or AI/methodology expertise
• Evaluation criteria:
  • Significance: potential contribution to measurement/analysis in the stated HPC primary research areas
  • Innovation: extent to which the AI approach advances beyond conventional methods
  • Feasibility: demonstrated relevant track record, technical and logistic feasibility within six months, and justified budget
  • Reproducibility and transparency: plans for code, documentation, bias assessment, and reproducible workflows where relevant
  • Potential for follow-on funding or broader impact (e.g., toolkit for other HPC Associates)

Notifications and reporting

• Decisions will be provided within approximately one month of submission.

• Final report: a final report must be submitted via an online form within 30 days of project end; HPC may request a short presentation or training session to share methods with HPC Associates.

Ethics, privacy, and responsible AI

Proposals should address ethical considerations where relevant, including e.g. not depositing micro data to commercial Large Language Models (LLM, e.g. ChatGPT), data privacy, consent, and de-identification. See the BSPH guideline as an illustrative example. When in doubt, the best practice is to submit an application to the IRB.

Projects involving human-subjects data must include IRB approval before funds release. Projects using sensitive or restricted data should explain data access and governance arrangements; HPC may require additional documentation before funds release.

Application and contact

Submit a single PDF containing the proposal, biosketch(es), and IRB documentation (if applicable) via the HPC MIF application form per current application procedures.

For questions, contact Dr. Wenxuan Huang: [email protected].