Artificial Intelligence In ICSID Arbitration: Is There A Need For A New Convention And Consent?
- Posted by: Research Team
- Category: Global Trade and Business Law
Ahmad Reza Mobini, (Ph.D. Candidate in Public International Law, Allameh Tabataba’i University)
The world is experiencing a technological evolution and artificial intelligence seems inevitable or better saying is the pivot of that. But what is artificial intelligence (“AI”)? It is beyond the scope of the present piece to undertake a detailed assessment of AI, but simply put, it is a complex data processing system provided with huge amount of data and algorithms. AI continues to affect nearly all aspects of human life ranging from industry to dispute settlement mechanisms, specifically arbitration.
Some have pointed to the opportunities that AI provides for arbitration. An example includes the role of AI in reducing the time and cost of arbitration processes. AI may also help arbitrators in some procedural tasks (e.g., translating and summarizing evidences) (see here). Another author not only mentions some advantages of AI, but also highlights some concerns such as the risk of hacking and distortion of the right to an explanation. Furthermore, another has described it as “not founded”, because the decisions that is made by such systems do not provide the arguments which usually are available in a decision taken by human arbitrators. Empathy is considered another important component which Frankman refers to (see here) and says arbitrating is a “cognitive exercise”.
Although there are some challenges and uncertainties, most experts have speculated that AI is likely to play an increasing role in arbitration at least in procedural issues. But we are a long way from having arbitrators replaced with robots. Yet, reflection on pros and cons of resorting to AI in arbitration, and possibly giving it the green light do not suffice to use this newcomer before discussing the issue of “consent”.
ICSID system as a dispute settlement mechanism sets forth a standard arbitration procedure. Whereas different investments do not share same characteristics and peculiarities, it seems that occasionally ICSID framework and its procedure are not sufficiently organized to cope with all of them. These specificities not only affect the way in which the arbitration is conducted but also push tribunals to adapt the available procedure, in order to “give effect to the choice of ICSID arbitration” (See Abaclat case (decision on jurisdiction and admissibility, para. 519). However, establishing an ICSID arbitration tribunal calls for a dual analysis; this must be undertaken according to the provisions of ICSID convention as well as the investment agreement governing the relationships between the host state and the investor. An important question that arises here is whether the state consent for arbitration in IA and the ICSID convention welcome such technology.
Generally, arbitration is defined as a dispute settlement mechanism which is totally consent-oriented. The arbitral tribunal must ascertain the consent of the parties to the referral of the dispute to an ICSID tribunal. The mere adoption of the convention by a state does not mean that it has consented to bring its disputes out of investments before the tribunal. A second consent is required. This second consent is materialized in three ways: in bilateral and multilateral investment agreements, national legislation or in a contract between the host state and the investor (see: “Investor-state arbitration”, Christopher F. Dugan & others, p. 220-221). The scope of the consent is limited to what is articulated in those instruments. That is a matter of treaty interpretation. Here it should be determined whether the application of artificial intelligence is a question of admissibility, jurisdiction or simply a matter of applying the ICSID rules and modalities. Having recourse to artificial intelligence in order to cope with procedural issues is not related to the question of whether there is an obstacle in the way of the dispute, nor is it associated with Article 25 of the ICSID Convention. In fact, it has nothing to do with the parties to the dispute and type of a dispute. In other words, as long as the instrument in question and Article 25 of the Convention allows specific parties to refer specific types of dispute to the tribunal, it doesn’t make any change whether or not the tribunal is going to resort to artificial intelligence.
As noted above, for the tribunal to be allowed to arbitrate the dispute, besides the ratification of the convention, the parties must have consented to the referral of a given dispute to that tribunal. None of these alone is enough for the tribunal to do its job. It must be recalled that the tribunal is limited to the contours of the consent articulated in the investment agreement. Likewise, the tribunal is not permitted to change the provisions of the convention; however, in the process of dispute settlement, the interpretation is inevitable.
Also, as discussed, the issue of artificial intelligence is a matter of Article 44 of the ICSID convention. This article reads as follow:
“Any arbitration proceeding shall be conducted in accordance with the provisions of this Section and, except as the parties otherwise agree, in accordance with the Arbitration Rules in effect on the date on which the parties consented to arbitration. If any question of procedure arises which is not covered by this Section or the Arbitration Rules or any rules agreed by the parties, the Tribunal shall decide the question.”
No doubt the ICSID articles and rules of procedures have not clearly prescribed use of such technologies, mainly because at that time artificial intelligence was not prevalent. But changes in the nature of investments (as is the case with mass claims) and emergence of some concerns (e.g., efficacy, time and cost-saving, transparency, predictability and …) required the adaptation of procedural rules parallel with those evolutions. A possible adaptation could be making a room for new technologies such as artificial intelligence.
In this regard the finding of tribunal in Suez and AWG v. Argentina (Order in Response to Transparency and Amicus Curiae Petition) is noteworthy; where the tribunal for the first time granted the amicus curiae the permission to participate in the proceedings. The tribunal held:
“The last sentence of Article 44 is a grant of residual power to the Tribunal to decide procedural questions not treated in the Convention itself or the rules applicable to a given dispute. (para. 10)” and before turning to consider the conditions under which the Tribunal may exercise that power the tribunal “concludes that Article 44 of the ICSID Convention grants it the power to admit amicus curiae submissions from suitable nonparties in appropriate cases (para. 16)”.
The same approach was followed by the tribunal in Abaclad case with regard to the mass claims. However, it is doubtful whether this power encompasses the possibility of transferring _even with regard to procedural issues_ the arbitrator’s job to a machine. That is obviously not an easy question to answer, so long as such change of incumbent bears legal consequences and policy implications for the parties as well as for the tribunal itself. Indeed, by consenting to the jurisdiction of a tribunal and referring the disputes to it, states look for a fully reasoned decision, an important and due demand that a machine is typically not expected to fulfill when it does not follow a cognitive process.
Concluding remarks: It might be argued that there is no need for a new form of consent in investment agreements. At the same time, however, it is difficult to reach an affirmative response with regard to the Convention. As mentioned, an examination of Article 44 of the ICSID Convention makes it possible to say with some certainty that precedent supports the capacity of convention for some adaptations. It is up to the tribunal to interpret the convention and its rules. But what is of paramount importance is that going too far would probably foment the backlash against investment arbitration. So, any change with any scope in the operating system of the international investment law would not succeed unless artificial intelligence and its benefits could encourage states to support such changes.