DSSG

What Makes a Strong Data Science for Social Good Project Partner?

The Data Science for Social Good (DSSG) Fellowship is a 12-week summer program launched in 2013 and built on more than a decade of learning and iteration. A successful summer depends on selecting the right problems and the right partners. Each year, we review hundreds of potential projects, scope them, evaluate data maturity, assess feasibility, negotiate data-use agreements, and determine whether a project can realistically produce measurable impact within a constrained timeframe.

Choosing well matters. Fellows need to make meaningful progress quickly, and project partners want to demonstrate real impact. Projects that are poorly scoped, lack sufficient data access, or do not have an engaged partner can stall momentum and reduce outcomes. Over the years, we have learned—sometimes through difficult experience—that strong partnership and operational readiness matter as much as technical sophistication.

This post outlines what we look for in a DSSG project partner.

Core Project Requirements

The following conditions are necessary (though not sufficient) for a successful fellowship project.

1. A Clearly Defined, Actionable Problem

We look for problems that are:

Concrete and decision-oriented.
The work should inform a real operational decision (e.g., who to prioritize for outreach, where to allocate resources, when to intervene).

Scoped to be feasible in 12 weeks.
We cannot solve systemic issues in a single summer, but we can target a well-defined decision within a larger system. DSSG cannot solve poverty in three months, but we can help reduce unemployment, homelessness, maternal mortality, lead poisoning, and school dropout rates by improving specific operational decisions within those systems.

Operationally embedded.
The output should integrate into an existing workflow, not sit on a shelf.

Projects that are purely exploratory, descriptive, or retrospective—without a clear path to decision-making—are generally not a good fit.

2. Meaningful Social Impact

We prioritize projects that:

  • Address high-stakes outcomes (e.g., health, housing, education, public safety, economic stability).

  • Affect clearly defined and meaningful populations.

  • Align with principles of equity and fairness.

  • Meet a real operational need within the partner organization.

We make a substantial investment in each project—financially, operationally, and opportunistically. When we choose one project, we forgo another. We therefore focus our limited resources on substantial problems with credible potential for impact.

We value projects that help larger populations over smaller ones and that address persistent, structural challenges rather than temporary issues. Past projects have spanned public health, human services, homelessness, education, economic development, international development, transportation, sustainability, disaster response, and environmental policy, though other domains may also qualify.

3. A Committed, Capable Partner

The strongest predictor of project success is the partner.

A strong DSSG partner:

  • Has deep subject-matter expertise.

  • Owns the operational decision the project will inform.

  • Has authority—or proximity to authority—to influence implementation.

  • Commits consistent time and attention throughout the summer.

Typically, this requires one to two individuals dedicating a meaningful portion of their time during the fellowship. Responsibilities include:

  • Pre-summer scoping.

  • A project kickoff presentation.

  • Weekly check-ins.

  • Iterative feedback on prototypes.

  • Post-summer evaluation, monitoring, and implementation planning.

Projects without sustained partner engagement rarely translate into real-world impact.

4. High-Quality, Accessible Data

Data maturity and access are often the most complex aspects of project setup.

Strong projects have:

  • Individual-level or sufficiently granular historical data tied to real decisions.

  • Clearly defined outcomes.

  • Reasonable documentation and institutional knowledge.

  • Legal and technical pathways for data access established before the summer begins.

Many DSSG projects involve sensitive data (e.g., health, education, justice). We operate within strict legal and ethical frameworks, and data-use agreements must be in place early. Delays in access can materially limit what is achievable in 12 weeks.

We work flexibly with partners—using de-identified data, background checks, secure remote servers, or other compliant arrangements—but we expect access to all relevant data necessary to solve the problem. Partial access often constrains impact.

All data are handled securely and confidentially in accordance with formal agreements.

5. A Path to Evaluation and Deployment

A technically successful model is not the goal. A successfully evaluated and used system that improves outcomes is.

We assess:

  • Where the output will live (e.g., case management system, batch file, API).

  • Who will use it.

  • How it will change behavior.

  • What program or policy intervention it will inform.

  • What constraints (policy, staffing, funding) affect deployment.

  • What IT and data infrastructure it will depend on.

  • How impact will be evaluated.

Projects without an evaluation plan and a plausible implementation pathway are unlikely to produce sustained impact.

6. Ethical, Legal, and Governance Readiness

Many DSSG projects involve high-stakes decisions affecting vulnerable populations. We evaluate:

  • Whether the decision is appropriate for the use of data, ML, or AI.

  • How performance, bias, and unintended consequences will be monitored.

  • What transparency is required.

  • Who remains accountable for decisions.

We treat AI systems as components of broader human decision-making systems—not as automated replacements for human judgment. Ongoing ethical and governance discussions are a core part of the fellowship, and we expect project partners to actively engage in them.

Fellowship-Level Considerations

In addition to evaluating individual projects, we consider how each project contributes to the broader fellowship experience.

Diversity of Domains and Methods

We seek a portfolio spanning multiple policy areas, methodological approaches, and institutional contexts. This diversity strengthens cross-team learning and demonstrates the range of data-driven impact.

Long-Term Partnership Potential

Sustained partnerships often produce deeper impact. Returning partners benefit from:

  • Established legal agreements.

  • Institutional knowledge.

  • Improved deployment readiness.

  • Iterative model refinement.

At the same time, we actively seek new partners addressing high-priority challenges.

Learning Value for Fellows

DSSG is both an impact initiative and an educational program. Strong projects:

  • Require interdisciplinary collaboration.

  • Involve real-world ambiguity and constraints.

  • Demand careful problem formulation.

  • Expose fellows to implementation and governance—not just model building.

Technically trivial or overly fragmented projects are not ideal.

What DSSG Is (and Is Not)

DSSG focuses on designing, building, validating, deploying, and monitoring data-driven decision-support systems for social problems.

We are a strong fit for projects involving:

  • Prediction, prioritization, classification, or optimization.

  • Causal inference tied to operational decisions.

  • Human–AI system design.

  • Evaluation of deployed models.

We are generally not a good fit for:

  • General data infrastructure development without an immediate decision target.

  • Website or application design.

  • Purely descriptive analytics without operational integration.

  • Long-term data collection efforts without existing usable data.

Organizations seeking broader data strategy support may be better served through longer-term collaborations outside the summer fellowship.

Why Partner with DSSG?

DSSG combines:

  • Technical rigor.

  • Operational grounding.

  • A focus on measurable impact.

  • Attention to governance and responsible AI deployment.

Our objective is not simply to build models, but to help organizations improve consequential decisions.

We view impact as a property of the full socio-technical system—data, model, workflow, incentives, and governance—not merely the code.

Interested in Partnering?

We review potential projects year-round. If you are interested in exploring a collaboration—whether through the fellowship or beyond—contact us at dssg@cmu.edu with:

  • A description of the operational decision you want to improve.

  • The population affected.

  • The data currently available.

  • The intended pathway to deployment.

  • The internal staff member who would serve as project lead.

Well-scoped, high-impact, implementation-ready projects with committed partners are the foundation of a successful DSSG summer.

How to get involved

We’re always looking for new ways to apply our skills for the social good. If you have a project for us to work on, please let us know. We look forward to hearing from you!