We present the methodology and the results of an application of argumentation theory to determine whether Radovan Karadžic´, President of the Serb Republic, possessed the mens rea-the knowledge of wrongdoing that constitutes part of a crime-for genocide in Srebrenica, where, in July 1995, at least 5,115 Bosnian Muslims were killed by members of the Serb Republic Forces. To evaluate the strengths and weaknesses of Trial Chamber's findings in the publicly available judgement, we used argumentation-based techniques available in the CISpaces.org tool. The results of our analysis were submitted to the Mechanism for International Criminal Tribunals (MICT) as an amicus curiaebrief, i.e., a brief from a non-party in a lawsuit who argues or presents information relevant to the lawsuit.
reported in [
The Trial Chamber’s finding that the accused possessed the mens rea—i.e. the intention
and knowledge of wrongdoing that constitutes part of a crime—for genocide in relation to the
Srebrenica joint criminal enterprise (JCE) was the subject of academic critique at the time of
the Trial judgement, e.g., [
Using the argumentation-based techniques available in the CISpaces.org tool [
Our main contribution is to show that the methodology we propose in Section 4 can be used to
show the weakness and strengths of a case—cf. Section 6. This can be of use for the plaintiff,
the defendant, but also judges and jurors as it helps clarifying which elements are beyond any
reasonable doubt, and which ones are not. This is currently a live issue in international criminal
law: one of the authors of this paper argues that “each piece of evidence should be evaluated on
its own merits, in light of the other evidence on the record, to determine whether a point has been
proven beyond reasonable doubt,” [
The submission of our amicus curiae triggered reactions from the academic community interested in international criminal justice, practitioners at the United Nations courts of law, and media. We critically analyse our research and comment on its impact and related work in Section 7.
What follows is a short historical summary of the events that lead to the Srebrenica killings as
reported in [
The Socialist Republic of Bosnia and Herzegovina (SRBiH) was one of the six republics that once constituted the Socialist Federal Republic of Yugoslavia (SFRY): unlike the other republics, it possessed no single majority ethnic grouping. One of its political parties, the Serbian Democratice Party or SDS—lead by Radovan Karadžic´, campaigns to establish separate Serbian institutions. Following a plebiscite held on 9 and 10 November 1991, the Serb Republic was proclaimed in 1992.
Among other key personnel within the Serb Republic, Radovan Karadžic´ served as President and Supreme Commander of the Bosnian Serb Army (VRS). Tomislav Kovacˇ was the Assistant Minister of the Ministry of Intern (MUP), and the acting Ministry from September 1993 until 4Dolus specialis: the specific intent to cause a specific kind of harm.
Tuzla River Drina
ITALY
CROATIA
January 1994. Ratko Mladic´ served as Commander of Main Staff, the highest operative body of the VRS. His assistant commander for Security Administration was Ljubiša Beara, with duties of management of the main staff of the Military Police, as well as co-ordinating with the bodies of the Ministry of the Interior. Momir Nikolic´ was Chief of the Security and Intelligence Organ, which was responsible for issues of security in the corps composing the VRS, including the arrest and detention of prisoners of war and other persons.
When in 1992 the population of Republic of Bosnia and Herzegovina voted for independence from the Socialist Federal Republic of Yugoslavia in a referendum, forces of the Serb Republic attacked different parts of the Republic of Bosnia and Herzegovina, whose state administration effectively ceased to function having lost control over the entire territory.The Assembly of the Serb Republic adopted the strategic goal to eliminate the border with Serbia: Srebrenica—a town with a majority of Bosnian Muslims—is close to that border (Figure 1).
In late June 1995, Karadžic´ gave a combat assignment that lead to an offensive against Srebrenica and ultimately to the killing of at least 5,115 Bosnian Muslim men.
Figure 2 summarises the timeline of the most relevant events for our analysis starting from the 6th of July. 11th, Afternoon • Srebrenica has fallen
Karadžic´ appoints Deronjic´ as Civilian Commissioner for
Srebrenica 11th, Night • A column of Bosnian Muslim men started moving in a
northwesterly direction towards Tuzla 12th, Morning • Shelling of the column began 12th, Afternoon • Large numbers of the members of the columns surrendered 13th, 1700h–1840h • Karadžic´ had an hour-long conversation on the phone during
Pale which he was briefed by General Maladic´ that Srebrenica “[wa]s
done.” 13th, Morning • 13th, 1630h–Night • Kravica Warehouse
13th, 2010h •
Intercepted call between Deronjic´ and
Karadžic´ through an intermediary
soldier and shot him dead: other soldiers started shooting at the detainees in response. Others shot at the detainees with machine guns and automatic rifles. Hand-grenades were thrown in the warehouse through the windows.
By nightfall, between 755 and 1,016 Bosnian Muslim men were killed.
: Deronjic´, the President is asking how many thousands?
For this analysis, we used the tools available from the CISpaces project [
Fundamental concept in computational models of argumentation is the one of inference rule, where a statement (antecedent) becomes a (prima facie) reason to believe another statement (consequent). For instance, “Mary, a witness, says that John committed the fraud” (antecedent) can be seen as a prima facie reason to believe that “John committed the fraud” (consequent). Some scholars distinguish between strict and defeasible rules in their approach to formal argumentation: as we will not make use of strict rules in this work, we will not discuss this further.
Rules provide the building blocks for the notion of argument, that—borrowing from the ASPIC
literature [
A statement is the contrary of another one when they cannot be both true, albeit they can both be false. Borrowing from the literature, a flexible way of using such a notion of contrariness is by allowing for a statement to be the contrary of another one. By requiring the vice versa, the two statements would become contradictory. We will make use of such a flexibility in the following of our analysis.
The notion of contrariness between statements leads to the concept of defeat between arguments: an argument defeats another argument if the former rebuts or undermines the latter. When the conclusion of an argument contradicts the conclusion of another argument, it is the case that the first rebuts the second, as well as all the other arguments that have such a second argument as sub-argument. If, instead, the conclusion of an argument contradicts one of the ordinary premises of another one, then the former undermines the latter.
Given a set of arguments and defeats between them, we need criteria to assess which arguments collectively survive the defeats and thus can provide a reasonable viewpoint (or extension) of the statements and the rules that we were considering. Such criteria usually consider conflict-freeness , i.e. the absence of defeats within the viewpoint; admissibility, i.e. if an argument in the viewpoint is defeated by a second argument, the latter must in turn be defeated by a third argument also in the viewpoint; and maximality, i.e. a viewpoint cannot be a strict subset of another viewpoint. Multiple viewpoints can exist for the same set of arguments and the defeats between them: two equally reliable witnesses, each providing one reason for contradictory conclusions, lead to the situation that each of the two arguments per se is a reasonable viewpoint, hence there are two of them. In this case, if an argument belongs to at least one viewpoint, it is said to be credulously accepted. If, instead, an argument belongs to all the viewpoints, it is said to be sceptically 5An interested reader can try the CISpaces.org at https://tiresia.unibs.it/cispaces/. accepted. In the following we will be making use of this notion of sceptically acceptance in connection with the principle of proof beyond a reasonable doubt.
CISpaces.org provides a convenient visual language and an effective Human-Machine Interface
for argumentation mapping. It builds on top of the Argument Interchange Format AIF [
Minor Premise: B has been observed to be true in a given instance.
Conclusion: Therefore, A is true.
CQ1: Is it the case that if A is true, then B is true? CQ2: Has B been observed to be true? CQ3: Could there be some reason why B is true, other than its being true because of A being true?
The other argumentation schemes used in this analysis are: the abductive argumentation
scheme; the argumentation from cause to effect; the argumentation from (popular) opinion; and
the argumentation from witness testimony [
Similarly to [
An argumentation system is as tuple AS = ⟨L , R, , ν⟩ where: • : L ↦→ 2L is a contrariness function s.t. if ϕ ∈ ψ and: ψ ∈/ ϕ, then ϕ is a contrary of ψ; ψ ∈ ϕ, then ϕ is a contradictory of ψ (ϕ = –ψ); • R = Rd ∪ Rs is a set of strict (Rs) and defeasible (Rd) inference rules such that Rd ∩ Rs = 0/; • ν : Rd ↦→ L , is a partial function.6
A knowledge base in an AS is a set Kn ∪ Kp = K ⊆ L ; {Kn, Kp} is a partition of K ; Kn contains axioms that cannot be attacked; Kp contains ordinary premises that can be attacked. An argumentation theory is a pair AT = ⟨AS, K ⟩.
To map a WDG into an ASPIC+ system, let us assume that: • P ⊆ N is the set of I-nodes, where each I-node in the graph is written pi; • ℓtype, with type = {Pro, Con}, refers to a S-node; • [p1, . . . , pn ↝ ℓPro ↝ pφ ] indicates an inference rule, where p1, . . . , pq are parent nodes of the S-node ℓPro, and pφ is a child of ℓPro; • conflict schemes can be either [p1 ↝ ℓCon ↦→ p2] or [p1, . . . , pn↝ ℓCon ↝ pφ ]. Given a WDG = ⟨N, ↝⟩, its corresponding ASPIC+ system AS = ⟨L , R, , ν⟩ is such that: • ∀p ∈ P ⊆ N, p ∈ L ; • R = Rs ∪ Rd with Rs = 0/ and ∀[p1, . . . , pn ↝ ℓPro ↝ pφ ], p1, . . . , pn ⇒ pφ ∈ Rd; • ∀[p1 ↝ ℓCon ↝ p2], p1 ∈ p2; • ∀[p1, . . . , pn ↝ ℓCon ↝ pφ ], is mapped as p1, . . . , pn ⇒ ph ∈ Rd and ph ∈ pφ ; and the knowledge base Kn ∪ Kp = K ⊆ L is such that, given [p1, . . . , pn ↝ ℓPro ↝ pφ ], ∀pi ∈ {p1, . . . , pn}, if pi is not a conclusion of any inference rule ̸ ∃[ℓPro ↝ pi] ∈ WDG, pi ∈ Kp. In addition, assume WDG′ = ⟨N′, ↝′⟩ a subset of WDG containing only a single cycle of inference schemes, then ∀pi ∈ P′ ⊂ N′ , if [ℓPro ↝ pi], [pi ↝ ℓPro] ∈ WDG′, then pi ∈ Kp.
Following [
6Informally, ν(r) is a wff in L which says that the defeasible rule r is applicable. However, we will not make use of this feature in the following.
An argument can be attacked in its premises (undermining) or its conclusion (rebuttal). Since
we will not use the preference ordering between arguments, we will omit it from the definition.
Similarly for the notion of undercut on the inference rule (cf. [
Given a and b arguments, a defeats b iff a successfully rebuts or successfully undermines b, where: a successfully rebuts b (on b′) iff Conc(a) ∈/ ϕ for some b′ ∈ Sub(b) of the form b′1′, . . . , b′n′ =⇒ –ϕ;a successfully undermines b (on ϕ) iff Conc(a) ∈/ ϕ, and ϕ ∈ Prem(b) ∩ Kp.
An argumentation framework (AF) [
An AF ⟨A , →⟩ is the abstract argumentation framework defined by AT = ⟨AS, K ⟩, AS = ⟨L , R, , ν⟩ if A is the smallest set of all finite arguments constructed from K (as above) ; and → is the defeat relation on A .
Given an AF ∆ = ⟨A , →⟩: a set S ⊆ A is a conflict-free set of ∆ if ∄ a1, a2 ∈ S s.t. a1 → a2; an argument a1 ∈ A is acceptable with respect to a set S ⊆ A of ∆ if ∀a2 ∈ A s.t. a2 → a1, ∃ a3 ∈ S s.t. a3 → a2; a set S ⊆ A is an admissible set of ∆ if S is a conflict-free set of ∆ and every element of S is acceptable with respect to S.
A set of argument S ⊆ A is a preferred extension iff it is a maximal (w.r.t. set inclusion) admissible set.
An argument is skeptically accepted w.r.t. preferred semantics iff it belongs to each preferred
extension. Checking this is a problem that lies at the second level of the polynomial hierarchy
[
The goal of our amicus curiae brief [
In the present case, Karadžic´’s mens rea is an element of the offence of genocide in Srebrenica,
as genocide requires each member of the joint criminal enterprise to be knowledgeable of the dolus
specialis of the principal perpetrator. The material facts upon which proof of mens rea hinged
were the Trial Chamber judgement’s findings on: (1) Karadži c´’s knowledge of the expansion of
the plan to remove Bosnian Muslims from Srebrenica to include the killing of men and boys,
hence Karadžic´ sharing the intent to destroy the Bosnian Muslims in Srebrenica; and (2) his
active involvement in the killings. Due to space limitations, let us consider only the former.
7In this paper we consider only finite sets of arguments: see [
Following [
It is worth noticing that the text proved resistant to attempts of automatic analysis. This is also evident in the graphical charting of our analysis (Figure 3), where we consider pieces of information spanning more than 210 pages (Para 5312 fn. 18025 [2, p. 2203] to Para 5808 [2, p. 2413]), in addition to historical information scattered around the entire document. For instance, the information that Srebrenica has fallen on 11 July 1995 has been presented in Para 5033 [2, p. 2079], 331 pages before being used in an argument to support the hypothesis.
Figure 3 depicts our understanding of the reasoning lines that the Trial Chamber describes in
its judgement [
There are three main lines of reasoning in favour of this conclusion. The first one is based upon Nikolic´’s testimony that he overheard Deronjic´ saying that the accused had instructed Deronjic´ that all detainees should be transferred to Zvornik, cf. Figure 2, 14th July 1995, just after midnight. This testimony gives reasons to the chamber to refute the alternative explanation—highlighted by the defence and reported in the judgement—that Karadžic´ was referring to a place different from Zvornik in the intercepted conversation with Deronjic´, cf. Figure 2, 13th July 1995, 2010h. In this line of reasoning, the chamber decided also to link additional pieces of information (Inference 3.cX of Figure 3), as supporting the conclusion that Karadžic´ ordered that detainees should be transferred to Zvornik, such as a complaint to Beara by Deronjic´ about the presence of detainees in Bratunac. However, for those facts, it seems to us that the chamber did not consider some relevant critical question, e.g. Is there any other reasonable explanation for why Deronjic´ had previously complained to Beara about the detainees’ presence in Bratunac, other than it being true because Karadžic´ conveyed to Deronjic´ the direction that the detainees should be transferred to Zvornik? Despite those additional pieces of information (Inference 3.cX of Figure 3), this line of reasoning does not rely on unstated findings or pieces of information for which critical questions have not explicitly been answered. Let us remind that we methodologically chose not to assess the reliability of Nikolic´ testimony as we did not have access to the entire historical records.
A second line of reasoning justifying the hypothesis is based on Simic´’s testimony that Deronjic´ told him that he had informed Karadžic´ about the events at the Kravica Warehouse the day after the incident, in conjunction with the unstated assumption that if the accused knew that Bosnian Unstated Generally, if the Accused knew that Bosnian Muslims had been recently killed by Bosnian Serb forces [in Kravica Warehouse] then he might have known that it may occur that Bosnian Serb forces would kill other Bosnian Muslim in the future
Cause to Effect Mladić informing Karadžić, on 13 July, that Srebrenica “[wa]s done” when Srebrenica had fallen on 11 July, hence Karadžić should have known by 13 July implies that Karadžić was knowledgeable of the intent to kill Bosnian Muslims
Abductive
Inference Karadžić was knowledgeable of the intent to kill Bosnian Muslims
Reasoning Line 3 On 13 July Mladić informed Karadžić that Srebrenica “[wa]s done” (Para 5768).
Srebrenica had fallen on 11 July, hence Karadžić should have known by 13 July (Para 5770, fn 19596)
Evidence to hyphothesis
Reasoning
Line 1 Inference 3.a Kovač gathered additional important information that he ultimately relayed back to the Accused when he returned to Pale on 14 July. (Para 5806) The Chamber finds it incredible that Kovač would not have discussed these matters with the Accused (Para 5782) Inference 3.b
Opinion
Opinion
Opinion The only reasonable inference is that Bajagić reported the events in Srebrenica he had witnessed [...] to the Accused during their meeting on 15
July. (Para 5783) Bajagić had substantive knowledge of the events in
Srebrenica (Para 5783) The extremely late hour of their meeting (Para 5783)
Abductive
Inference The Accused had invited Bajagić to Pale (Para 5783)
Evidence to hyphothesis During the second meeting [with the Srebrenica representatives, ed.], Deronjić reported on the situation in Srebrenica (Para 5808)
CON Meeting Bajagić in Pale at an extremely late hour, given that Bajagić had substantive knowledge of the events in Srebrenica implies that Bajagić reported the events he had witnessed Unstated
Beara and Deronjić later argued about whether the detainees would be killed in Bratunac or would be transferred to Zvornik for that purpose (Para 5773)
Deronjić [...] [said] that the Accused had instructed him that all detainees should be transferred to Zvornik (Para 5773) Witness Testimony
Nikolić testimony (Para 5312, fn 18025)
Witness
Testimony
Muslims had been recently killed by Bosnian Serb forces (in Kravica Warehouse), then he might
have been known that it may occur that Bosnian Serb forces would kill other Bosnian Muslims
in the future. Similarly as above, we have some doubts that all relevant critical questions find
an answer in the judgement [
A third line of reasoning is based on an abductive inference with the unstated premise that Mladic´ informed Karadžic´, on 13 July, that Srebrenica “[wa]s done.” The Trial Chamber appears to have concluded that, given that Srebrenica had fallen on 11 July, Karadžic´ would have known this by 13 July. From that unstated inference, it drew a further inference that the conversation implied that Karadžic´ knew of the intent to kill the Bosnian Muslims of Srebrenica.
On 21 February 2018, we sent to the United Nations Mechanism for International Criminal
Tribunals (MICT) a request for leave to make submissions as amicus curiae pursuant to Rule 83
of the MICT Rules of Procedure and Evidence [
Given the overall results (cf. Section 5), we believed that our analysis was particularly useful for the prosecutor in the case, as it highlights a few ways to strengthen the case. Much to our surprise, on 5 March 2018 the prosecution replied to our request asking the Appeals Chamber to deny the motion because amicus curiae submissions are generally limited to questions of law, and not to questions of facts. On 9 March we then replied that whilst our amicus curiae brief focuses on the question of Karadžic´’s mens rea for genocide in Srebrenica, it goes to the heart of an important legal question on the interpretation of the standard of proof, i.e., whether each piece of evidence should be evaluated on its own merit or whether it is acceptable for the Trial Chambers to find their decision on the basis of the accumulation of all the evidence in the case but without the need to link factual and legal findings to the final decision.
With an even greater surprise, on 12 March, the Defence Counsel asked the Appeals Chamber to accept the proposed submissions and give the parties the opportunity to make their own observations on those submissions, adding that “[t]he Proposed Submissions bring judicial fact-finding into the 21st century.”
This is clearly not the first attempt to bring “judicial fact-finding into the 21st century.” In [ 17], Verheij introduces the notion of automated argument assistance which is in spirit very close to our work here, as it explicitly aims at drafting and testing of court pleadings. Walton [18] provides an extensive account of argumentation in legal systems, and argues for a new method for legal argumentation which, among other, includes the use of “argument diagramming to map out the network of inference in a given case” [18, p. 323]. For completeness of discussion, authoritative colleagues criticise the use of argument diagramming, notably van Gelder in [19]. In reflecting on his experience, he notices that argument diagramming might not serve well the purpose of deliberation, possibly because deliberation is a dialectical activity rich in nuances. However, he did not consider deliberation activities where the incentives for a proper epistemic investigation are significant, such as writing a judgement for an international criminal case.
The work by Walton on legal argumentation [18] and in general on argumentation schemes—
summarised in [
We also feel that there is very little we can add to van Gelder observations in [26], where he analyses some legal arguments. His comments strongly resonate with us, as we also experienced “little use of verbal indicators of logical structure, and often use obscure or vague indicators,” [26] incomplete arguments, with also text scattered across the document, and possibly serving multiple purposes.8 Although far from providing an off-the-shelf support tool, in retrospective our analysis would have significantly benefit from (1) entity-relations extractors, e.g. [ 27] and (2) topic modelling system, e.g. [28], which together might transform a static PDF document into a database that can be queried. As it is evident in the graphical charting of our analysis (Figure 3), we considered pieces of information scattered across more than 210 pages (Para 5312 fn. 18025 [2, p. 2203] to Para 5808 [2, p. 2413]), in addition to historical information written in different part of the document: for instance, the information that Srebrenica has fallen on 11 July 1995 has been presented in Para 5033 [2, p. 2079], 331 pages before being used in an argument.
Differently from previous approaches, we considered a case under discussion at ICTY offering the results of our analysis as an amicus curiae brief to the Appeal Chambers. It unfortunately denied admissibility of our application on 28 March 2018, observing that “the issues regarding whether Karadžic´ the mens rea for genocide in relation to the Srebrenica JCE were extensively litigated before the Trial Chamber and have been fully briefed by Karadžic´ and the Prosecution on appeal.” The Appeals Chamber also seems to criticise the fact that “the Amicus Curiae Observations seek to guide the Appeals Chamber’s analysis of the Trial Judgement without consideration of or access to the entire record that is relevant to the Trial Chamber’s conclusions.” This however would raise the question: what is the purpose of having a 2615 page judgement, if the judgment does not actually fully reflect the grounds for the conclusion? Finally, confirmation that our analysis was trustworthy comes from the Appeal Chamber Judgement [29] that in its section D.2 provides a summary of the Trial’s Chamber Judgement regarding whether Karadžic´ was knowledgeable of the intent to kill Bosnian Muslims which is almost entirely present in our 8A reviewer of an earlier version of this paper commented that some of the instance of argument from opinion in Figure 3 seem more of instance of argument from ignorance. This is something that only a judge mindful of the purpose of their prose could clarify when writing the document. resulting argumentation network depicted in Figure 3.
In this paper, we present the methodology and the results of an application of argumentation
theory to determine whether Karadžic´ possessed mens rea for genocide in Srebrenica based on
[
Although at first sight this paper seems to be similar to other attempts to analyse legal reasoning
with formal argumentation, e.g. [22, 23], it differs from them substantially as we did not try to
capture the debate component, hence distinguishing between Prosecutor and Defence claims.
Instead, our analysis is closer to works analysing arguments in a single document, like, for
instance, [
Although the Appeals Chamber denied the admissibility of our application, the interest that applying formal argumentation theories triggered in the international criminal law community suggests that there is scope for future work in this area. We cannot claim that the methodology used in this analysis is beyond critique, but we can claim that it can help creating a better judgement that fully reflects the grounds for the overall conclusion.
This is the long-term aspiration of the ongoing research underpinning this paper, and we are fully aware that this will require to provide answers and innovative proposals both from a technical perspective as well as from the legal one. From a technical perspective, for instance, we still lack appropriate methodologies for adequately transforming statements of natural language into formal logic—a problem most students of logic encounter without being presented with satisfactory solutions, cf. among others [30]—thus inevitably exposing the subjectivity of each formalisation. In addition, following [31], we will also work in the direction of assessing the quality and the strengths of different argumentation reasoning lines, by taking into consideration quantitative measurements of uncertainty and trust, thus enriching the community proposals looking at probabilistic elements in legal reasoning, e.g. [32, 33, 34].
This research formed part of a project entitled Developing a Tool to Support Judicial Reasoning in International Criminal Trials, funded in part by Cherish-DE, the United Kingdom’s Digital Economy Crucible (http://cherish-de.uk).
This research was partially sponsored by the Italian Ministry of Research through a Rita Levi-Montalcini Personal Fellowship (D.M. n. 285, 29/03/2019) and by the U.S. Army Research Laboratory and the U.K. Ministry of Defence under Agreement Number W911NF-16-3-0001. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the U.S. Army Research Laboratory, the U.S. Government, the U.K. Ministry of Defence or the U.K. Government. The U.S. and U.K. Governments are authorized to reproduce and distribute reprints for Government purposes notwithstanding any copyright notation hereon. [17] B. Verheij, Bart, Automated argument assistance for lawyers, in: ICAIL ’99, New York,
New York, USA, 1999, pp. 43–52. [18] D. N. Walton, Legal argumentation and evidence, Pennsylvania State University Press, 2002. [19] T. van Gelder, Cultivating Deliberation for Democracy, Journal of Public Deliberation 8 (2012). [20] F. Bex, H. Prakken, C. Reed, D. Walton, Towards a Formal Account of Reasoning about Evidence: Argumentation Schemes and Generalisations, Artificial Intelligence and Law 11 (2003) 125–165. [21] F. Bex, S. van den Braak, H. van Oostendorp, H. Prakken, B. Verheij, G. Vreeswijk, Sensemaking software for crime investigation: how to combine stories and arguments?, Law, Probability and Risk 6 (2007) 145–168. [22] H. Prakken, Formalising ordinary legal disputes: a case study, Artificial Intelligence and
Law 16 (2008) 333–359. [23] H. Prakken, A. Wyner, T. Bench-Capon, K. Atkinson, A formalization of argumentation schemes for legal case-based reasoning in ASPIC+, Journal of Logic and Computation 25 (2015) 1141–1166. [24] L. Al-Abdulkarim, K. Atkinson, T. J. M. Bench-Capon, A methodology for designing systems to reason with legal cases using abstract dialectical frameworks, Artificial Intelligence and Law 24 (2016) 1–49. [25] K. Atkinson, T. J. M. Bench-Capon, Relating the ANGELIC methodology and ASPIC+, in:
COMMA 2018, 2018, pp. 109–116. [26] T. van Gelder, Why are legal arguments to hard to follow?, https://timvangelder.com/2009/ 08/13/why-are-legal-arguments-so-hard-to-follow/ (on 23 Aug 2018), 2009. [27] M. Surdeanu, D. McClosky, M. R. Smith, A. Gusev, C. D. Manning, Customizing an information extraction system to a new domain, in: ACL2011, 2011, pp. 2–10. [28] H. Zhuge, L. He, Automatic maintenance of category hierarchy, Future Generation
Computer Systems 67 (2017) 1 – 12. [29] ICTY, Prosecutor v. Karadžic´: Appeal Judgement, MICT-13-55-0660/2, 2019. [30] M. Baumgartner, T. Lampert, Adequate formalization, Synthese 164 (2008) 93–115. [31] F. Cerutti, G. Pearson, Supporting scientific enquiry with uncertain sources, in: 21st
International Conference on Information Fusion, 2018, pp. 404–411. [32] H. Prakken, Argument schemes for discussing Bayesian modellings of complex criminal cases, in: JURIX 2017, 2017, pp. 69–78. [33] C. S. Vlek, H. Prakken, S. Renooij, B. Verheij, A method for explaining Bayesian networks for legal evidence with scenarios, Artificial Intelligence and Law 24 (2016) 285–324. [34] B. Verheij, F. Bex, S. T. Timmer, C. S. Vlek, J.-J. C. Meyer, S. Renooij, H. Prakken, Arguments, scenarios and probabilities: connections between three normative frameworks for evidential reasoning, Law, Probability and Risk 15 (2016) 35–70.