Investigating decision accountability and trustworthiness with dominance-based rough sets analysis

During the covid-19 pandemic, several countries including the UK, took the approach of tiered restrictions - making separate decisions on the level of restriction of movement to be imposed in different geographical areas - as opposed to a single country-wide decision on restriction of movement. The UK government proposed a set of five factors to determine the separate level of restrictions to be imposed in each area. Although such tiered restriction systems appeared to be a more sensible approach, many found them unsatisfactory, due to a lack of sufficient transparency, affecting trust and acceptance of such decisions. In this work, we explored the fairness of such allocation decisions by using the Dominance-based Rough Set Approach (DRSA). DRSA is invariably utilised to create a single set of decision rules from a (single) dataset to analysis for insights and to utilise on unseen data. In this work, we instead utilise separate data segments, pertaining to separate geographical areas, to create separate DRSA rule sets for each. Then, we compared the separate rule sets in terms of their consistency and fairness. We found inconsistencies that highlighted an overall trend of discrepancies between the North and the South of England. Through drilling down further, we found that the inconsistencies were driven largely by London, suggesting that London was treated more leniently than the rest of the country. The notion of comparing consistency between different DRSA rule sets could be utilised within a plethora of other domains, to explore decision consistency, helping to enhance decision trustworthiness and accountability