Embed Notice

For the record, the actual problem domain for me is climate change actions for cities, like putting in bike lanes, electrifying the bus fleet, or having a heating oil to heat pump rebate. My team made a heuristic comparator for actions (reduced emissions/cost, applicability for city bioregion, side benefits) but it only gets about 70% right on comparisons compared to human experts. We also trained an ML model based on the expert feedback, and that does better but not great.