Governments can also help stimulate areas of measurement where there is existing work, but a need to improve tools and establish more continuous measurement. For example, when it comes to assessing fairness in AI systems, there are existing datasets and measures, but not as many as might be needed for fine-grained evaluation of new AI capabilities. In this case, governments may seek to proactively create datasets (rather than use existing ones) that make it easy to test a system for a highly specific form of fairness (e.g. audio speech datasets of different regional accents to test how well speech recognition approaches can serve different accents).
