Back in 2019, I wrote about a paper by Joost Pauwelyn and Krzysztof Pelc on the role of the Secretariat in WTO disputes, following up on Joost's own post about the paper. As Joost put it, this is what they did and what they found:
We begin by comparing the textual similarity of rulings on the basis of their potential authors. We then collect external texts written by all the potential authors of WTO rulings to create stylometric profiles for each. Using computational authorship-attribution tools, we then apply these profiles to determine who the more likely author of a given ruling is. The findings are unambiguous: the WTO Secretariat appears to exert significantly more influence over the writing of WTO panel reports than the adjudicators themselves.
Joost and Krzysztof have continued their work on this issue, and have two EJIL blog posts, an EJIL article (free version here), and an AJIL article. Their EJIL article also discusses the identification of authors of dissents:
The WTO is the only major international tribunal to allow dissenting opinions but require that they be anonymous. The ostensible purpose of anonymity is to provide cover from political retribution to individual adjudicators wishing to express a minority opinion. ... We use text analysis to confirm just how thin this veil of anonymity currently is. To demonstrate the point empirically, we analyse the most recent AB dissenting opinion, issued in a politically sensitive case brought by China against the United States, US—Countervailing Measures. Using external texts by all the candidate authors, we build a stylometric profile for each one, and compare it to the text of the opinion, which allows us to identify the likely author of the dissent—the Appellate Body member from the United States—with a great level of certainty. Our intent in doing so is merely to demonstrate that given the current design of three-member-divisions, the veil of de jure anonymity is insufficient to protect the identity of individual opinions from the scrutiny of Member-states.
Starting with the role of the Secretariat, now that it is more widely known -- and has been clearly established with evidence rather than just something insiders talk about -- that the Secretariat folks play such an important role in writing panel reports, Joost and Krzysztof have offered some suggestions for reforms that will maintain the appropriate "equilibrium." For example, they propose "disclosing the names of staff involved in a proceeding and circulating the 'issues papers' and any econometric advice provided by Secretariat staff to the parties to a dispute, and then publishing these memos alongside the final ruling, would prove significant"; and they suggest that "[t]erm limits could be imposed on the number of years that specific individuals can work for the Legal Affairs or Rules Divisions or for the AB Secretariat."
While I'm curious what governments and others think of these suggestions, for now I'm thinking more about the reaction to Joost and Krzysztof putting this information out there. They have shown that, in many cases, software can be used to identify the authors of WTO dispute rulings. How will these authors react?
If I were a Secretariat official assisting on these cases, I would consider it important not to be identifiable as the writer of an opinion. The point of that job is to assist the actual adjudicators. In doing so, it's possible that the Secretariat could have some substantive influence here and there. Maybe on occasion a Secretariat official would come up with a good argument that the adjudicator adopts. In my view, though, if I as a Secretariat official am so involved that my writing style comes through in the ruling, I'm probably playing too big a role and I should step back a bit. If I were asked to draft part of a ruling, my goal would be to emulate the voice of the person who asked me to write it, and if software demonstrated that my own voice was coming through, I would make some adjustments.
Turning to adjudicators writing dissents, the issues here are slightly different. The rules may require anonymous opinions, but it is not clear that anonymity can always be preserved. In most AB dissents, I think many close observers of WTO dispute settlement can make a pretty good guess without the benefit of software (panel dissents are probably more difficult). If it were me in this position, I might try not to use too personal a style, and my dissent from a majority opinion would not read like, say, my blog posts. But other people may be less worried about this, and thus they wouldn't make efforts to avoid being detected by software or just people's intuition.
So what will happen now? With Joost and Krzysztof having demonstrated the capability of the software, and people at the WTO Secretariat now on notice that their writing is being scrutinized, how will the officials who work on WTO disputes react? We may find out soon whether the authors of these rulings have the same view as I do. Will they adjust their writing? Will the WTO purchase a copy of the text analysis software to help authors avoid detection? In three years or so, will Joost and Krzysztof still be able to identify the authors of WTO rulings? I encourage them to revisit the issue and see if there have been changes in the capability of their software.
Regardless of what happens here, the role of the Secretariat is an important question and we should discuss it, and I'm curious what WTO Members think. What might be helpful is if they had a better understanding of the Secretariat's role at present. Some government officials already have a sense of this because they have worked on cases, but others may not. A survey-based study of what goes on behind the scenes, and the range of approaches that panelists and the Secretariat take, could be useful here.
