A society that can learn how to improve people’s lives.

Consider a neighborhood organization with an idea for helping older residents remain connected to the people around them. Its staff know the local circumstances. Residents have views about what would be useful. There is enough support to begin, but a great deal remains uncertain: who would take part, what would make the activity worthwhile, and whether it would change the isolation the organization hopes to address.

In the future we want, that uncertainty has somewhere to go. The organization can obtain competent help, examine related work, involve residents in the questions, and choose a proportionate way to learn. It can discover that part of the idea needs changing while there is still time to change it. A useful finding can reach another community with enough explanation for people there to judge its relevance.

The same possibility belongs in public administration. A school district considering a different way to support attendance should be able to learn from experience elsewhere and investigate the conditions in its own schools. A public agency changing how it delivers benefits should be able to examine who finds the new process easier and who encounters a new obstacle. Serious questions should have access to serious attention.

More room to attempt something worthwhile

When inquiry is within reach, an idea can develop before its advocates have assembled an expensive case for it. An uncertain beginning can receive patient examination. People can test a practical change, discover why it fell short, and try again with a clearer understanding of the problem. Public work gains a greater capacity for creation as well as correction.

This requires institutions that can respond constructively to what they learn. A result may justify expansion, expose a need for redesign, or suggest that resources would do more good elsewhere. The people responsible for those choices need both an intelligible account and the authority to act on it. Those affected need a way to be heard while the choices are still open.

Learning should also survive the circumstances that produced it. A project ends, staff move on, a government changes. The account of what was tried, why, and under which conditions can remain available. Successors should be able to examine a decision without having to reconstruct its reasoning from the people who happen to remember it.

Across generations, that continuity could widen the range of undertakings societies can sustain. Communities could cooperate around a shared uncertainty. Institutions could revise responsibilities that no longer fit the problem. A new organization could begin with the benefit of earlier attempts, including the unsuccessful ones.

This is a future worth working toward: a greater ability to improve lives through purposeful experimentation, accumulated understanding, and institutions capable of learning. Reaching it requires attention to the conditions that make learning difficult now.

Institutions already learn in many ways. Public servants use research and administrative data. Practitioners notice when a service is failing. Communities explain what an official account has missed. Researchers, evaluators, and independent reviewers produce knowledge that improves public work. The ambition begins with that experience and asks how its benefits can become more widely available.

One obstacle is the cost of bringing the right attention to a question. A record may span years of reports, changing definitions, interviews, spreadsheets, and decisions made by people who have left. Understanding it requires someone to establish what happened and what each source can support. The effort needed to prepare an inquiry can consume the resources that would have paid for a better answer.

Another obstacle is that information is collected for particular purposes. A service record helps staff deliver a service. A grant report accounts for agreed activities. Financial controls establish whether money was spent appropriately. Each has value. Asking whether an undertaking improved people’s lives introduces further questions about outcomes, causes, distribution, and cost.

A record can be complete for its original purpose and insufficient for that later question. An attendance count says who arrived; understanding who could not attend requires a different source. A rise in employment among participants describes a change; estimating the program’s contribution requires a credible account of what would otherwise have happened.

The institutional conditions matter

Even where a question can be answered, the institution may struggle to use the answer. A study can finish after a budget decision. Its findings can cross several departments while responsibility for responding belongs to none. A manager can inherit an indicator whose definition no longer fits the service. These difficulties concern how work is organized as much as how analysis is performed.

Incentives can make the problem harder. A funder may want honest learning while asking each organization to demonstrate success at renewal. A program leader may welcome a better account of results while fearing that uncertainty will be treated as evidence of failure. An evaluator may depend on repeat commissions from the institution whose claims are being examined.

These tensions can arise among people acting carefully within their responsibilities. They call for arrangements that make honest inquiry usable and survivable: questions agreed before results arrive, resources for learning, authority to report inconvenient findings, and decisions that distinguish a weak claim from the worth of the people doing the work.

Some disagreements persist after the facts are clear. An institution can know that a change would help and still lack the authority, money, or political support to make it. People can agree about an effect and disagree about whether its costs are acceptable. Evidence must enter those decisions without pretending to settle every consideration within them.

The practical task is therefore larger than commissioning another study. It includes making inquiry affordable, connecting it to responsibilities and timing, and preserving room for findings that change the expected account. That is the setting in which new technical capabilities become consequential.

For the institutional dimensions of evidence use: OECD, Building Capacity for Evidence-Informed Policy-Making.

There is something remarkable in being able to ask a machine to read a body of work, compare accounts within it, reason through a question, and return to relevant material as the inquiry develops. Tasks that once depended entirely on a person’s available attention can now receive substantial machine assistance.

These capabilities reach material that institutions have often struggled to use fully: reports that need to be read together, open responses that need careful examination, a change in policy buried among successive versions, or research scattered across unfamiliar fields. A model can help identify patterns, propose explanations, and expose a question that deserves closer investigation.

The significance extends beyond producing the same report more quickly. When the effort required to explore a question falls, more questions become practical to pursue. A small organization can consider inquiry earlier. An expert can enter a discussion with more of the background already assembled. A promising idea can receive attention before a large research budget exists.

Attention can reach further

Imagine a public agency preparing to change a service that several other jurisdictions have already tried to improve. The relevant knowledge might include evaluations, implementation notes, public consultations, and accounts of changes abandoned along the way. AI can assist with finding and comparing that material, connecting statements to sources, and identifying differences requiring professional attention.

The expert’s work can then concentrate more fully on what matters locally: which comparisons deserve weight, what the agency can deliver, whose experience is missing, and which uncertainty could change the decision. Better preparation can also reveal that existing knowledge is sufficient, or that a proposed study would answer the wrong question.

This opening reaches the design of new inquiry too. Systems can assist with reviewing instruments, drafting research materials, checking consistency, and maintaining the record as a project changes. Savings in that supporting work can make a stronger design feasible. The actual gain depends on the task, the quality of the material, and the work required to check the output.

Dependability has to be built

A model can miss a decisive passage, misread a table, or produce an explanation unsupported by its sources. Its fluency can make those errors difficult to notice. Access to a large volume of text is a capacity to examine it; dependable coverage requires deliberate checks. Research on long contexts has shown why the placement of relevant material can matter to retrieval.

Useful machine reasoning therefore belongs within work that can be inspected and challenged. Sources need to remain accessible. Important statements need checking. The question and the limits of the record need to stay visible as the work develops. People must retain authority to reject the proposed explanation and pursue another one.

We see a substantial opportunity in building those conditions around expanding model capabilities. Greater access to inquiry can give people more room to create, adapt, and learn together. The next chapter concerns the standards that let that opportunity become useful knowledge.

Research on capability and limits: retrieval-assisted literature synthesis and language models’ use of long contexts. These studies examine particular systems and tasks.

A useful inquiry starts by identifying what someone needs to understand. Whether a service reaches the intended people, why implementation varies, how participants experience it, and whether it causes an improvement are distinct questions. They can require different evidence and lead to different decisions.

Existing research may answer part of the question. Administrative records may describe delivery. Interviews may reveal a barrier no standard measure captures. A new study may be warranted where an important uncertainty remains. The method follows the question, the circumstances, and the strength of the conclusion the decision requires.

Causal questions need particular care. To estimate a program’s effect, an inquiry must establish a defensible comparison with what would have happened without it. Random assignment can provide that comparison in appropriate settings. Other designs use variation in eligibility, timing, or circumstances, each with assumptions that need examination. A model’s prediction acquires causal meaning only through a justified design and assumptions.

Technical sophistication cannot restore information that was never recorded or remove uncertainty merely by redescribing it. Sometimes the responsible conclusion is that a proposed question cannot be answered well enough from the available material. That conclusion can still guide a decision about what to say now, what to change, or what would be worth learning next.

Knowledge accumulates with its boundaries

Learning that an approach helped in one setting creates a further question about where else it might help. Staffing, institutions, local knowledge, the people involved, and the surrounding services can affect what happens. The conditions of a successful study belong in the account that travels with it.

As people examine a finding across settings, they can develop a richer understanding of the circumstances in which it holds. A difference between two studies can become useful evidence about implementation or context. Keeping those differences visible gives future inquiry more to build on than a list of favored interventions.

Qualitative and quantitative work each contribute to this understanding. An estimate can describe a pattern that interviews help explain. Experience can identify an outcome missing from the instrument. A careful inquiry makes clear what each form of evidence contributes and where professional interpretations differ.

Systems that support demanding work

Prova’s proprietary evidence systems organize inquiry around questions, claims, sources, and decisions. They direct AI-assisted work, preserve connections to the record, and make gaps, conflicts, and unresolved questions visible for review. Experts can examine the reasoning, challenge assumptions, and keep uncertainty present through the conclusion.

Named people remain responsible for consequential judgment. They must understand the context, consider alternatives, and explain why the evidence supports the conclusion reached. A reader should be able to distinguish what was observed, what was inferred, and what the responsible institution decided to do.

Cheaper inquiry can extend this discipline to more of public life. The same opportunity makes persuasive but poorly supported conclusions easier to produce. Source connections, independent challenge where needed, and accountable interpretation become more valuable as the volume of analysis grows.

Further reading on choosing and planning evaluation: HM Treasury’s Magenta Book.

The ability to ask more questions gives institutions more power over the people described in their records. Used well, it can reveal exclusion, correct an unfair account, and help a service respond to experience it previously overlooked. Its legitimacy depends on the terms under which that knowledge is sought and used.

Three commitments have equal standing in the future we want to help build. Each can change the question, the design of the inquiry, and what may responsibly be done with the answer.

Affordability includes the burden on people

Analysis can become less expensive while the demands placed on everyone else increase. Another survey still takes time. Linking records can create new exposure. A request that seems modest in one project may be the latest in a series of requests to the same household or frontline team.

The cost of inquiry should therefore include the time, disruption, and risk borne by those who provide its information. Existing material should be used where appropriate, new collection should have a clear purpose, and the value of another answer should justify what obtaining it asks of people.

Imagine that a public servant with a consequential question could obtain competent evidence support as an ordinary part of the job. The help would begin with the decision and the available knowledge. It could draw on local experience, relevant research, administrative material, and specialists where the question required them. Its scale would fit the stakes.

Now imagine that the account remained useful afterward. A successor could understand why a choice was made. Another institution could examine the approach. A community could challenge a public claim. With appropriate permissions and protections, a finding could travel with enough of its context to support further inquiry.

Together, those capabilities would change the starting conditions of public work. An institution would have access to help in the present and a usable inheritance from work already undertaken. Each new project could contribute to that inheritance.

A stronger basis for cooperation

Organizations do not have to agree on an entire strategy to investigate a question together. A public agency, a nonprofit, and a community group might begin with a common concern about who is being missed by a service. They could agree on how to examine it, preserve their different interpretations, and use the findings to identify a practical next step.

That kind of cooperation could make new undertakings possible. Partners could establish what they need to learn before committing to a permanent institution. Several funders could support inquiry whose value extends beyond any one portfolio. Public bodies could reconsider a division of responsibilities when evidence shows that the problem crosses the boundaries within which they operate.

Institutional renewal would still require authority, resources, public deliberation, and choices about competing purposes. Better knowledge gives those processes more to work with. It can clarify a disagreement, reveal an assumption that no longer holds, or show that a proposed arrangement deserves a fair trial.

Learning that moves in more than one direction

The ambition extends across countries and communities. India and countries beyond the OECD belong within it as contributors of methods, experience, and institutional invention. Knowledge should be able to move in several directions, with local professionals and communities examining what another setting’s findings mean for their own.

Translation involves more than language. A practice can depend on relationships, public authority, working conditions, or a form of trust that a report barely mentions. Reciprocal learning takes those differences seriously and gives the people receiving an idea a role in judging and adapting it.

Public infrastructure can take several forms

A capability this widely available could be supported by public agencies, independent institutions, professional networks, universities, commercial providers, or combinations of them. Its durability would depend on access, financing, standards, maintenance, and governance. Which functions belong together and who should own them are questions to work through.

Prova’s contribution begins with practical work that can help establish what such a capability requires. Real questions expose where assistance is useful, where expertise remains essential, and what can responsibly be carried from one setting to another. That is how a firm working today can contribute to a future extending far beyond itself.

The future described here would be built by people working in many different settings: public servants, practitioners, researchers, communities, funders, and people creating institutions that do not yet exist. Each brings knowledge and responsibilities that the others cannot supply alone.

Making inquiry affordable is one part of that work. Making it trustworthy and useful is another. Keeping it available through changes of leadership, protecting room for disagreement, and ensuring that people can build on what is learned require sustained institutional effort.

We want Prova to contribute to those conditions through work people can examine and use. The questions encountered in practice should improve the systems, sharpen the methods, and inform the larger ambition. Others may find better approaches, identify consequences we have overlooked, or build complementary and independent versions of the capability.

That openness matters to the purpose. A public capability gains value when people can contribute to it, question its standards, and carry useful knowledge into circumstances its originators never anticipated. Its future should remain responsive to what those people learn and choose.

If this is work you want to follow, support, participate in, or challenge, we welcome a conversation. You may bring a practical question, methodological experience, an institutional opportunity, or a different account of what society needs. Each can help give the ambition a more useful form.

We begin with a possibility worth pursuing: that people trying to improve lives could have much greater access to the knowledge and help their work deserves, and leave more behind for those who follow them.