Ce document est lié à :
info:eu-repo/semantics/altIdentifier/doi/10.1007/s13563-019-00206-2
Fenintsoa Andriamasinoro et al., « Use of artificial intelligence to assess mineral substance criticality in the French market: the example of cobalt », HAL-SHS : économie et finance, ID : 10.1007/s13563-019-00206-2
The French public and commercial stakeholders need prospective tools to follow how mineral substances criticality change in the French market. After arguing that such tools should necessarily tackle criticality at a complex level, in particular on multiple scales (e.g., France and the EU), we present the first thematic and methodological discussions of our results from the ongoing design of a methodologically based simulation model on two subfields of artificial intelligence: agent-based computational economics (ACE) and machine learning (ML). In applying this to cobalt, our model aims to assess a supply shortage in France for prospective purposes. More precisely, we model a first individual agent (which is already complex by itself) acting at a country level: France. This model is not yet an ACE model per se since only one agent is designed. Nonetheless, we include ACE in the discussions since the work is a premise of such an end. The discussions also include how well the field accepts the methodology. At a thematic level, our preliminary prospective conclusion is a French cobalt supply shortage, should the case arise, would not be due to the variation of price from the UK, the transit leader of cobalt export to France. At a methodological level, we think the idea of methodologically coupling ML and ACE is necessary. ML is well-known in this field, but mainly for the study of mineral prospectivity in mining. Conversely, ACE covers the value chain but is not yet well known in the field and as such is still not trusted.