# Learned survey policy digs where the University's surveyors always have

> A research poster from the School of Underlying Conditions trains an active-learning policy to map soil boundaries and finds that its pits settle into the same pattern as the survey's own archived field sheets.

A machine-learning policy asked only to draw the lines between soils more accurately has learned to put its pits where the University's soil surveyors have been putting theirs for years. That is the central result of _Where the Policy Digs_, a new research poster from the School of Underlying Conditions by Dr Brynjar Obiora, Senior Research Fellow, and Dr Bram Ntuli, Lecturer and Convenor of Improvement Grand Rounds.

> Operations research tends to earn its keep by telling a field something it half knew already. In this case the field was a paddock, and what it half knew was where to put the spade.
>
> --- Dr Bram Ntuli, Lecturer and Convenor, Improvement Grand Rounds

The study sets a trained sampling policy, a set of evenly spreading statistical designs and the survey's archived pit records side by side across forty simulated landscapes and one unnamed survey area, and reads each resulting arrangement of pits through the lens of how its density fluctuates from window to window. Inside a mapped soil the policy spreads out almost as evenly as a lattice. Near the line between two soils, it crowds in.

The poster takes its place in the University's growing body of work where computation meets the ground beneath it, and adds a new instrument to the School's long-running conversation about where one soil ends and the next begins.

"Every pit is a small decision about a line," said Dr Obiora. "It seems those decisions travel rather well."

The full poster is available from the University's research repository under an open licence, doi:10.5555/slop.9ho8qe.

Source: <https://slop.university/news/2026-10-19-learned-survey-policy-digs-where-surveyors-always-have/>\
Licence: [CC-BY-4.0](https://creativecommons.org/licenses/by/4.0/)\
Cite as: Swift, B. (2026). _Slop University_ [Artwork]. https://doi.org/10.5281/zenodo.21199776
