Research
Indicators that improve and show their work
How BIT is building a continuing process of methodological challenge, validation, and improvement with Forge.
An indicator is not just a number. An indicator is a piece of information that reflects choices about what to measure, which evidence to trust, how to handle gaps, and which uncertainties matter. As new evidence becomes available, those choices should be open to examination.
At BIT, we are developing Forge to make that examination a continuing part of our work. Forge connects methodologies, evidence, code, and review decisions so that improvements can be investigated systematically. It helps us retain what we learn, identify unresolved questions, and test promising approaches against the methods already in use.
Consider a familiar research decision: whether a newly identified dataset would strengthen an existing indicator. In our urban heat exposure and resilience work, Forge has helped identify and document recently published datasets as candidates for comparison. The new dataset’s relevance makes it worth investigating, but differences in observation periods, spatial resolution, and definitions of urban areas need to be assessed. Each dataset remains a candidate until an assessment establishes how it can contribute.
This is a small example of a broader strategic position. Access to more research, more data, and more computational capacity expands what we can investigate. Turning that capacity into dependable information, indicators and insights requires a disciplined process for deciding which changes deserve adoption—and under what conditions.
Forge is being built around that process. Proposed code changes are challenged against our methodological standards. Research records distinguish exploratory findings from accepted decisions. Overnight reviews help surface evidence gaps and opportunities for further improvement. Together, these capabilities create a more continuous connection between research and implementation.
Our approach also allows useful development and analysis work to proceed while uncertainty remains. A promising method can be explored before it is ready for production. A working implementation can be tested before its outputs are empirically validated. The responsibility is to make those stages explicit and require evidence appropriate to the claims being made.
That changes how we think about an indicator methodology. We want it to be an executable research process: connected to its sources, clear about its assumptions, and capable of revision when a better approach earns its place.
AI and machine learning methods give us greater capacity to search, compare, implement, and investigate. BIT analysts and our rigorous reviewers remain responsible for the scientific questions and consequential judgements. Forge helps make the supporting work systematic, repeatable, and available to the next person who needs to examine it.
Self-improvement, in this sense, means building the conditions under which improvement can be demonstrated. Our ambition is straightforward: indicators that remain useful as evidence evolves, supported by methodologies that can improve—and explain why.