In December, SPARC assessed an institutional agreement that a Dutch national academic consortia and Elsevier were in the process of negotiating. At the time, we were responding to leaks in the press, which were largely confirmed by the subsequent release of the terms of a framework agreement between the Dutch consortia and the publisher. Last week, the parties announced the official terms of the agreement.
As a quick recap, we originally noted five concerns:
- Danger of linking publishing and data contracts into a “Bigger Deal”
- A deal structure inhibiting competition in data analytics services
- The implications of the resulting reduced competition on customer leverage
- The creation of a monopoly (or quasi-monopoly) on data analytics resulting in the loss of diversity in academic assessment
- The risks that the deal’s structure, if replicated, would pose to the overall health of the scholarly publishing ecosystem
While some new details have emerged since SPARC released our initial analysis, none of them materially change our conclusions. The newly released FAQ page on the deal 2 only adds to our concerns. The language of the FAQ is couched in ways that appear reassuring (“There is no exclusive relationship and third parties can also co-develop services and tools”), but the reality is that the academic institutions who signed the deal may have little incentive to use multiple vendors providing similar services. Having more than one tool to perform any evaluation only adds costs and complexity and is likely to be economically unsustainable.
The only area of concern articulated in our original analysis where we still have to suspend judgement is the potential impact of the deal on the economics of scholarly publishing, because the full financial terms of the agreement have not been made available. We cannot confirm if the Dutch consortia really received a zero increase on the aggregate “Read and Publish” costs in exchange for introducing Elsevier’s data analytics products (as the original leak suggested). In the absence of confirmation, we cannot determine for certain whether the deal leads to a path that may hurt smaller publishers, affecting the overall health of scholarly publishing – but the absence of specifics is, in itself, one of the many problems posed by the actual deal.
Sarah de Rijcke, the Scientific Director of The Centre for Science and Technology Studies at Leiden University, an institution dedicated to studying scientific research and its connections to technology, innovation, and society, raised a number of troubling questions in a recent blog post. Professor de Rijcke notes:
-
The agreement allows Elsevier to build the modules that would represent the building blocks of a national Dutch Open Knowledge Base and potentially undermines plans for a community-owned initiative in this area. Professor de Rijcke notes that the current deal structure provides Elsevier what amounts to “an insurmountable competitive advantage in terms of access to research intelligence.”
-
The agreement raises the question whether Elsevier is requested to deposit proper metadata to Crossref, and whether this metadata (including citations) will be fully open.
-
The money spent in acquiring Elsevier tools now becomes unavailable to support competing services; a general tender to all interested third parties would have been a better approach.
Professor de Rijcke concludes “I am not persuaded by the contract, and still find disconcerting that this deal may effectively transfer crucial means to influence Dutch science policy to a monopolistic private enterprise.”
In addition to Professor de Rijcke’s concerns, there are three additional elements of the agreement to take note of.
-
Principles. In early 2020, the participating consortia charged a Task Force with establishing terms and conditions under which public research metadata could be used by public and private organizations, and with outlining the development of an open knowledge infrastructure in cooperation with private vendors. As part of this work, the Task Force formulated guiding principles to define the relationships between academic institutions and third-party organizations in the release and use of metadata.
These principles are still in a period of open consultation, but it appears that they were used to drive the Elsevier agreement. The principles identified by the Task Force are certainly worthwhile (they state, for example, that ownership of content and metadata must reside with the institutions and researchers, rather than with third-party vendors; interoperability must be provided, etc.), but they are incomplete.
A number of additional important principles should have been included. The Berkman Klein Center at Harvard University recently released an analysis of Ethical and Rights-Based approaches to using AI5 aggregated from a survey of 36 prominent AI principle documents; yet almost none of the themes covered in this survey are reflected in the Task Force guidelines. Many of the items excluded should be viewed as non-negotiable principles prominently featured in any long-term relationship with a third-party vendor and translated into proper contractual requirements. These include:
-
Areas of exclusion. No mention is made of areas of activity that should be precluded from third-party vendors (for example, ranking of departments across academic institutions or faculty evaluation and assessment for the purpose of advancement). The Elsevier agreement does not list any activities or purposes that are explicitly barred.
-
Strong privacy protection. European data regulation (GDPR) provides significant privacy protection, but it would be nonetheless helpful to expand on some of its provisions for the sake of clarity. For example, it would be strongly advisable for faculty members and researchers to:
-
Give their consent to the collection of data only after a detailed explanation of how it will be used, and any changes to this use should also be clearly explained.
-
Have a right to restrict data usage to a specific list of tasks and should not be forced to provide a blanket acceptance. There should be no consequences from the decision to opt out partially or completely.
-
Have a broad right to demand erasure, as well as rectification of data that is wrong or incomplete.
-
Within applicable laws, receive notification of data requests from any government or government agency; the data only handed over in the presence of a legitimate court order.
-
Be immediately notified of any data breach.
-
-
Algorithm transparency. Strong preference should be given to vendors using Open Source algorithms. When this preference cannot be fulfilled, there should be rights of inspection by an independent third-party which can audit algorithms for errors and biases and report its findings to users.
-
Equity. There should be clear evidence that algorithms are developed in line with the overarching goals of the academic community (for example, sustainability and proper diversity among the designers of the algorithms) and that there are proper mechanisms for the correction of identified errors and biases.
-
-
Contractual transparency. Non-disclosure clauses remain a consistent problem. One of the mechanisms that has made it more difficult for the academic community to manage negotiations with commercial vendors has been the consistent lack of transparency on pricing, terms, and conditions. In this case, only the total annual amount of the contract from 2020 to 2024 is disclosed, with no breakdown among separate services and no comparable data for spending in 2019. While the Dutch consortia are within their rights to maintain de facto non-disclosure of the financial agreements signed within this agreement, there are consequences for taking this course of action: other academic institutions and consortia around the world will continue to negotiate at a disadvantage due to information asymmetry.
-
Conflict of Interest. One of the inescapable contradictions of asking a publisher to also involve itself in analysis that affects academic decisions is the risk that those analyses will favor authors that publish in its journals. It is unfair to ask Dutch researchers to not take into consideration, when they debate where to submit their next article, that the funding of their next project may be decided with tools operated by a company that is also in the business of publishing articles. Nothing in the agreement (or in the draft guiding principles which were used to sign the agreement) seems to make any mention of this evident conflict of interest.
This agreement raises significant issues both in terms of some of the novel problems it poses, such as the questions raised by Professor de Rijcke around the impact on competition in the development and provision of data analytics services, as well as more familiar ones around contractual transparency and conflict of interest. We hope that institutions exercise caution before entering similar deals, and follow approaches aimed at minimizing the negative consequences of signing “Bigger Deals.”
