Proving Genocide: A Follow-up to Marko Milanovic

In his recent – as always highly inspiring – post on proving genocide, Marko Milanovic makes five points on the “correct legal position” of which I will challenge here the second and fifth one.

1. First of all, Marko correctly states that the only reasonable inference (ORI) standard is not limited to circumstantial evidence regarding genocidal intent but applies generally to inferences based on circumstantial evidence. Indeed, the Kosovo Specialist Chambers provide for a specific rule in that respect (Rule 140(3) of its RPE), and its Appeals Panel (AP) referred several times to the ORI standard in its recent Shala judgement (paras. 662, 687, 689, 692, 699, 701, 703, 706, 709, 713, 716, 757, 789 and 815), holding inter alia:

The Appeals Panel therefore finds that a reasonable trier of fact could have concluded, on the basis of the totality of the circumstantial evidence, that the only reasonable inference was that the Murder Victim was deprived of his procedural guarantees … (para. 692)

In light of this evidence of a consistent pattern of a lack of procedural guarantees at the KMF [a detention camp] … the Appeals Panel is satisfied that a reasonable trier of fact could have determined, on the basis of the totality of the circumstantial evidence, that the only reasonable inference was that they were deprived of their procedural guarantees. (para. 713)

2. However, contrary to Marko’s claim (in his second point), there is no necessary link between ORI as an inference standard and beyond any reasonable doubt (BARD) as a standard of proof (which, noted in passing, is also applied in civil law jurisdictions, if only under a different name – “intime conviction”, “Überzeugung” – see recently from a comparative perspective Ambos/Lee, in Ambos et al., Core Concepts III, 2025, pp. 263 ff.). In fact, these are two different standards: namely one (the inference standard) qualifying the (logical) process of drawing inferences from circumstantial evidence, i.e., describing a process of deduction (see here); and the other one (the standard of proof) defining the necessary standard for the evaluation of evidence with a view to a certain procedural/judicial stage. Thus, we are dealing here with distinct intellectual processes: inferring vs. evaluating. There are different standards of proof depending on what needs to be proven and at which procedural stage: a prima facie case (initial suspicion) to start preliminary investigations, a “reasonable basis” to open a formal investigation (Art. 15(3), 53(1) ICC Statute), “reasonable grounds to believe” to issue an arrest warrant (Art. 58(1)(a) ICC Statute) or “substantial grounds to believe” to confirm charges (Art. 61(5) ICC Statute). There may also be different inference standards, e.g. an only possible inference standard (arguably even stricter than ORI) or a simply reasonable inference standard (without “only” and thus arguably more flexible than ORI [I note in passing that the focus on reasonableness is the gist of the more flexible approach in the Canada et al. submission, paras. 51-52, in the Gambia v. Myanmar ICJ proceedings]), but these inference standards are independent of the applicable standard of proof. It is therefore at least an over-generalisation if Marko argues: “ORI is directly linked to BARD, and if BARD … is not the relevant standard of proof then requiring ORI is not appropriate either.” This view also overlooks that a finding BARD even for a decision on guilt is not required for each fact to be proven (Bemba et al. Appeal Judgment, para. 868), which might render the ORI standard unnecessary but not “inappropriate”.

Concretely speaking, if, for example, a judge or chamber decides on the confirmation of a genocide charge, she will apply the respective standard of proof (“substantial grounds to believe”) with regard to ORI as the inference standard for proving genocidal intent through circumstantial evidence. Thus, the chamber may ask whether the available evidence, meeting the substantial grounds to believe standard, allows an inference of genocidal intent as the only reasonable inference. Alternatively, the chamber may ask whether there are substantial grounds to believe that genocidal intent may be the only reasonable inference based on the available evidence, i.e., alternative inferences may still be reasonable (and admissible due to this lower standard of proof as compared to BARD). Thus, while the standard of proof may have an impact on the inference standard (here, re genocide, ORI), these standards are, at any rate, still different in the sense explained above.

I note in passing that the case law ­– part of which I have analysed in an earlier post – does not address the issue of the distinction between a standard of proof (BARD) and an inference standard (ORI); it rather confuses or ignores this distinction. A good example is the ICTR Appeals Chamber judgement in Zigiranyirazo: while discussing ORI in one para. (49) in connection with the burden of proof, the Chamber just makes the (fairly obvious) point that the ORI standard does not apply to the defence since it does not bear the burden of proof (fn. 136). Later case law is equally beside the point. The Al Bashir ICC Appeals Chamber, para. 33, makes exactly the doctrinal mistake I am arguing against: it conflates a standard of proof (BARD) with a standard of inference (ORI) by stating that “the only reasonable conclusion amounts to requiring … to eliminate any reasonable doubt.” The recent Yekatom/Ngaïssona Trial Judgment (para. 136) does not deal with the issue but just states the obvious, i.e., that a trial chamber can draw inferences based on circumstantial evidence (as the KSC AP quoted above) and that the BARD standard does not prevent that (in fact, it does not speak to that at all). The case is nevertheless illuminating with regard to the difference between ORI and BARD: in the face of large amounts of (especially digital) evidence, chambers are increasingly required to evaluate the evidence holistically at the end of the trial, as did the Trial Chamber in Yekatom/Ngaïssona (in more detail here and here). While this holistic evaluation is being made based on BARD, the “findings on the facts and circumstances described in the charge” that are “reached by way of inference” are a prior, separate step (para. 661).  

3. If the aforesaid is correct, i.e. ORI applies, in principle, to the inference to genocide based on circumstantial evidence independent of the applicable standard of proof, this should have consequences on how other (non-judicial) institutions, international organisations or States/governments determine whether genocide has been committed (in Gaza or elsewhere). Thus, I would add a further nuance to Marco’s fifth point where he argues that “other fact-finders are not bound to apply a high standard of proof or ORI, for genocide or other international crimes alike.”

First of all, this statement is overly broad and overinclusive treating all fact-finders equally. One surely needs to distinguish between institutionalised fact-finding mechanisms, like a UN Commission of Inquiry, and purely private mechanisms, like an NGO, let alone assessments by individual academics and/or citizens. Secondly, if these other fact-finders were not bound or not even, at least, guided by the judicially accepted standard (or any standard), the genocide charge could be levelled arbitrarily by anybody turning a primarily legal claim into a purely political one, using “genocide” – as in fact often happens in public discourse – as a synonym for an especially grave and serious conduct and resulting harm (thus in fact often equating it with widespread and serious war crimes and/or crimes against humanity).

4. Thus, one should differentiate as follows: as to institutionalised fact-finding mechanisms and official actors (governments/States, international organisations) the same distinction between a standard of proof and an inference standard applies, albeit mutatis mutandis (taking into account especially that such mechanisms may not have the same robust institutional constraints, e.g. rules of evidence, as a State judicial system). It is therefore, in principle, correct when the UN Commission of Inquiry on the Occupied Palestinian Territory applies the ORI standard (paras. 159-60) to its much lower standard of proof, namely “reasonably grounds to conclude” (para. 7). It does that, however, noted in passing, somewhat confusingly and rather affirmatively than actually applying and reasonably explaining the standard (see esp. para. 220: “Commission finds…”; crit. also Talmon). Of course, ORI, while “settled jurisprudence” (Ntagerura et al., Appeals Judgment of 7 July 2006, para. 306), is a demanding standard (as can be seen in the 2016 Karadzic judgment, see here), perhaps too demanding (Canada et al., para. 51).

As to private actors, a further differentiation is necessary. One can surely require more from an NGO, especially a global one, than from ordinary citizens. As to the former, in principle, the same (judicial) standard should apply if only for the fact that (global) NGOs claim a quasi-judicial authority when they accuse States/governments of serious human rights violations or international crimes. In fact, serious NGOs apply that standard anyway, as the very discussion on a possible genocide in Gaza shows (see e.g. Amnesty International, pp. 97, 102 ff.; Human Rights Watch, pp. 29, 162-3, 167; Physicians for Human Rights Israel, paras. 107 ff.; B’Tselem, pp. 12-3 [but generally adopting a broader understanding of genocide]). In contrast, one cannot require the same methodological rigor from simple citizens. Indeed, ordinary common-sense reasoning does not, as Bentham famously argued, employ formalistic constraints on logical inference. But here too a minimum of rationality and rationalisation can be expected and legal / judicial standards may help to achieve that. After all, we all bear epistemic and moral responsibility for any claims we make and this responsibility increases with the (moral) gravity of the claim.

5. At the same time, applying the correct and generally recognized legal standard, does not mean that one cannot make a legal assessment (in Gaza) before a final authoritative judgment (by the ICJ) has been issued. In fact, States are obliged to undertake such an assessment in order to comply with their prevention and repression obligations under the Genocide Convention. The prevention obligation is triggered when a State “learns of, or should normally have learned of, the existence of a serious risk that genocide will be committed” (ICJ, Bosnia and Herzegovina v. Serbia and Montenegro, Judgment of 26 February 2007, para. 431). As to Gaza this has arguably happened with the ICJ’s first provisional measures order of 26 January 2024 in South Africa v. Israel, and the respective risk awareness should have increased by the since then aggravating situation on the ground (cf. ComInq, para. 249; for the ensuing awareness of the Netherlands de Hoon and Sluiter; conc. Sexton).

6. Last but not least, it needs, again contra Marko, to be stressed that genocide is special but not so much for the evidential standards to be applied (the inference standard and the standard of proof) or an alleged abstract hierarchy between international crimes but for its legal elements (actus reus and mens rea). In essence, these aim, on an objective level, to protect a certain group against its destruction and, on the subjective level, require the specific or special genocidal intent (dolus specialis, dolus directus first degree). As a result, if we say – as government representatives, members of an NGO or even ordinary citizens – that a genocide is taking place in Gaza, we all are saying – or we mean to say (in legal terms) – that Israel (state responsibility) and/or its official representatives (individual criminal responsibility) is/are attacking the protected Palestinian group with a special intent to destroy it, this special intent being the only reasonable inference from the available circumstantial evidence. This could well be the case, as argued elsewhere (see here and further developed here), but we should not mischaracterize the offence definition, or water down the evidentiary standards.

I thank Dr. Alexander Heinze and Prof. Paul Roberts as well as the editors for valuable comments.