A.I. for Impact

The four dimensions of a whitebox. Environmental, social and governance factors are the main factors deciding a company’s or an investment’s sustainability footprint and subsequent risks.

The box is white because it is open: every value carries its source, method and time of recording, every model shows which inputs drove its output, and the people described can look in, check and reply.

A community knows the water has changed before a satellite does; a workforce knows a site is unsafe before an auditor arrives. Every whitebox therefore begins with engagement: what the stakeholders a project touches know, and what they say back as things change. Public and private datasets, disclosures, news, satellite observation, sensors and market data come after, each checked against what the people on the ground said and never in place of it.

The sources a whitebox is built from

The data in a whitebox is large, mixed and always changing. No person can read all of it; machine learning can. The models read across the dimensions at once: the risk building in one place from signals in another, the direction a situation will take, the anomaly that means fraud or waste, the opportunity for positive impact that nobody has priced. Two rules hold. A model must show its working, since a black box inside a whitebox defeats the purpose, and its output is a reading to be checked by the people it describes, not a verdict.

Monitoring and reporting: how the A.I. is put to work

The cycle is hard to see because its halves live in different reports: a benefit to a community is a social reading, the permit it protects a governance one, the delay it prevents a financial one. Held together through time, the four dimensions let a model trace an impact to the risk it lowers, and a mitigated risk to the impact it raises.

Most sustainability reporting is a rating assigned once a year, from outside, after the fact. A whitebox reads from inside and in time: impact is measured where it lands, a risk to the environment shows as the social, governance and financial risk it also is, and a change is caught while it is still a trend. A company can manage impact and risk instead of reporting on them afterwards, an investor sees a project as it is, and a community can hold a project to its word with the project’s own record.