Artificial intelligence for robust multicriteria analysis in mine closure option assessment

Authors: Hari, E Download Paper Open access courtesy of:

DOI https://doi.org/10.36487/ACG_repo/2615_118

Cite As:
Hari, E 2026, 'Artificial intelligence for robust multicriteria analysis in mine closure option assessment', in AB Fourie, G Boggs, J Heyes & M Tibbett (eds), Mine Closure 2026: Proceedings of the 19th International Conference on Mine Closure, Australian Centre for Geomechanics, Perth, pp. 1-15, https://doi.org/10.36487/ACG_repo/2615_118



Abstract:
Multicriteria analysis (MCA) is widely used in mine closure planning to compare closure options across competing objectives such as regulatory compliance, environmental performance, social value and technical effectiveness. When applied well, MCA provides a clear and defensible basis for decision-making. However, long-term closure decisions are often characterised by uncertainty in site conditions, diverse stakeholder perspectives and sensitivity to assumptions that are difficult to fully represent using fixed weights and single point scores alone. This paper presents an artificial intelligence- (AI) enhanced MCA workflow that strengthens established closure MCA practice by introducing preference learning, stochastic evaluation and explainability. Preference learning captures criteria weights from practical pairwise comparisons across 5 stakeholder groups, allowing different perspectives to be represented transparently rather than collapsed into a single negotiated weight set. Stochastic multicriteria acceptability analysis is used to evaluate option performance across many plausible futures by sampling from uncertainty ranges in scores and, where applicable, weights. Explainability tools based on Shapley value decomposition highlight which criteria most strongly drive rankings and where trade-offs are most significant. A conceptual case study demonstrates the workflow applied to a legacy closure problem with 3 options assessed against 6 criteria under 3 climate scenarios. Under the regulator perspective and a dry climate scenario, the preferred option achieved rank one in 91% of simulated futures, providing a clear and quantified measure of robustness. The approach is implemented in Excel with AI computation support and retains professional judgement, established governance and existing workflows throughout.

Keywords: artificial intelligence, multicriteria analysis, mine closure planning, decision support, uncertainty, explainability

References: