In the deep level, the relationships between entities (i.e. scores) are represented by link values, that is the presence/absence of a link (deterministic sub-model) and the likelihood of a link (Bayesian sub-model). On that basis, analysts may elicit in parallel different working hypotheses. In the working level, these link values are materialised or not, depending on each relevant hypothesis (WHa, WHb, WHc, WHn) formulated by operators in the forensic intelligence process (e.g. forensic scientists, intelli-gence analysts or investigators). In Fig. 4, WHa considers that the link detected between grey and white entities according to the deterministic threshold is not significant. WHb considers the links detected by the deterministic approach but disregards the potential link between grey and black entities supported by a limited or moderate likelihood ratio. WHc considers a link between grey and black entities that was not detected according to the deterministic threshold that is a link based on a moderate or limited likelihood ratio or based on the indirect link between grey and black entities. Analysis can be initiated based on these working hypotheses. At this stage, other forensic science information or alternative information may be leveraged to assess the relevance of each hypothesis. Thus, the boundary between the evaluation and analysis steps is not that clear within the forensic intelligence process [4].