 |
Pix Credit 1902 Cartoon of a more ancient form of corruption; HERE
|
Since the Enlightenment and the rise of narratives of quantitative divinity in the West, it has become common to deepen cultural presumptions that (1) numbers do not lie; (2) that data serves as its own defense against corruption; (3) and that "following the science" inevitably serves the community as protection against the corruption of discretionary governance by humans. As a consequence cultures grounded in metrics, in quantitative assessment, and in the fortune telling of predictive analytics has come to dominate an administrative culture now rebuilt to satisfy collective desires to manage virtually every aspect of human life. That management, itself is a function of the mathematics of self-reflexive measure of collective action by reference to its own statistics, and related to that, by using this self-reflexivity (we measure ourselves for and by reference to ourselves and in that way identify deviance and promote conformity with the core) to shift the center of desired "bell curves" in the "right" direction. Ratings systems do that--establishing hierarchies of aggregations of measurable activity against a desired ideal. A related mathematics is applied to the science of accountability, of measuring performance against ideals or (better put) expectations; or even better put of conduct relative to that of others ((a bell curve without an anchor in a defining ideal, but one that produces substantial pressure to change behavior to conform to the center (or in hierarchical modalities in which measurement is meant to push aggregated behaviors of the community) toward the right side cluster of outliers meant eventually to define the center of the curve.
But the language and meaning making projects of mathematics is as corruptible as any other human activity. That corruption can infect the entire process of algorithmic governance (even in its form as ratings systems based nudging and embedding of values) is hardly surprising. That it can compromise the entire process of quantification--from the determination of the data to be collected, to the preservation of its legitimacy (issues of data robustness), to the development and application of analytics, and to the way that the rules are developed and applied to add judgment to the results of the exercise of analytics on data--ought to be less surprising still. The stakes can be quite high--and the size of the stakes suggests the central importance of algorithmic governance as an important site for creating facts, for disciplining behavior, and for managing the overarching structures of rules (law) through which the administrative governance of quantitative management is undertaken. And always at the center of these forces are the human elements and their institutional manifestations--individuals, internal and external institutional systems, and the competitions among individuals and institutions (including states) within managed systems of hierarchical status based authority.
To manage algorithmic governance one must first manage the people who tend to or control such systems. To manage people one returns to the ancient problem of the socialization of context-appropriate societal values--or in their absence, aggressive systems of accountability grounded in the principle that it must be presumed that people cannot be trusted. And to meet the challenge of context specific socialization one returns (again) to the naturalization of values and the construction of societal inducements to conform to those values (a problem of social credit in China and of data driven nudging rewards and punishments though markets in the West). All of this is a work in progress with resistance as strong as the societal forces that appear to be pushing human collectives in the direction of well managed values enhancing systems of quantified data based governance.
One thing is clear--such layered and multiple systems will not come cheap and they will be as complex as the underlying model of algorithmic governance (including modeling and predictive analytics). Another is that the business of data driven government is as messy as they of making law.In both cases the human element--the constant negotiation on several levels--personal, institutional, trans-institutional, social, and transnational--all operate to create a sociology of quantitative metrics and their analytics that is every bit as susceptible to politics (understood in this context sometimes as the corruption of the algorithm, though it merely points to its ultimate failure as another product of humanity striving to overcome its own humanity) as traditional bureaucratic regulations and the lawmaking among bodies of elected officials.
On 16 September, the World Bank Group distributed a statement (World Bank Group to Discontinue Doing Business Report) admitting "data irregularities on Doing Business 2018 and 2020" that resulted in the "initiat[ion of] a series of reviews and audits of the report and its methodology" as well the the activation of "the Bank’s appropriate internal accountability mechanisms." As a consequence of the invocation of these accountability and audit mechanisms, and on the basis of a Report submitted by members of the Law firm of Wilmer Culter Pickering Hale & Dorr, "the Board of Executive Directors, World Bank Group management has taken the decision to discontinue the Doing Business report." A Statement on Release of Investigation into Data Irregularities in Doing Business 2018 and 2020 then followed.
Today, the World Bank provides an excellent example of both the importance and corruptibility of quantitative systems, and of their repercussions. But at the same time, the World Bank example suggests the development of those metrics--of those qualitative as well as quantitative measures-- that are being operationalized, at least in some sort of conceptual form, as a toolkit of those quality control (accountability) measures now necessary to reduce (to tolerable levels) systemic corruption in a way that parallels the use of the process of administrative oversight in law based systems. Nonetheless, the ancient problem remains--the corruption of the human individual at the center of the effort.
The WilmerHale Report got to the heart of the issues that the World Bank thought too indelicate to more clearly specify in its statements:
1. On January 20, 2021, the International Bank for Reconstruction and Development (“the Bank”) engaged our firm to review the internal circumstances at the Bank that contributed to the data irregularities identified in the Doing Business 2018 and Doing Business 2020 reports. To that end, we undertook to understand: (1) how improper changes to the data for China (Doing Business 2018) and Saudi Arabia, the United Arab Emirates, and Azerbaijan (Doing Business 2020) were effected; (2) who at the Bank directed, implemented, or knew about the changes to the data and how their direction or pressure manifested; and (3) what internal circumstances, whether related to policies, personnel, or culture, allowed for the changes to take place.
2. Building on investigative work undertaken by the Bank’s Office of Ethics and Business Conduct (“EBC”), our review has proceeded on two investigative tracks simultaneously: The first focused on the ethical aspects of conduct relating to Board officials and was conducted pursuant to the Code of Conduct for Board Officials; the second focused on potential misconduct by Bank staff members and was conducted pursuant to applicable Staff Rules and Directives. . . Over the course of our investigation, our team collected roughly five million documents from Bank employees; reviewed 80,000 of those documents most likely to contain relevant information; and interviewed more than three dozen current and former Bank employees.(WilmerHale Report ¶¶ 1-2).
The factual conclusions included politically sensitive conclusions:
27. The changes to China’s data in Doing Business 2018 appear to be the product of two distinct types of pressure applied by Bank leadership on the Doing Business team: (1) pressure—both direct and indirect—applied by senior staff in the Office of the President, presumably at the direction of President Kim, to change the report’s methodology in an effort to boost China’s score; and (2) pressure applied by CEO Georgieva and her advisor, Mr. Djankov, to make specific changes to China’s data points in an effort to increase its ranking at precisely the same time the country was expected to play a key role in the Bank’s capital increase campaign.
* * *
33. The evidence demonstrated that the changes to Saudi Arabia’s and UAE’s data in Doing Business 2020 was likely the result of efforts by a senior Bank staff member to achieve a desired outcome and reward Saudi Arabia for the important role it played in the Bank community, including its significant and ongoing RAS projects. (Ibid).
And these were followed by a series of recommendations respecting the World Bank's compliance cultures, their auditing and accountability systems, and the structures for the protection of the integrity of data based governance systems.
 |
Pix Credit (Zambia Billboard HERE)
|
While it is tempting to find irony here--one of the global leaders for the development of governance capacity itself found it unable to deploy sufficient capacity to avoid significant corruption of its own data based systems. Nonetheless, it is also possible to see in this the way that systems will have to continuously (re)develop capacity as they invest more and more in quantitative governance undertaken by large and complex aggregations of human inputs. What data driven governance reminds us is that even the most sophisticated systems of algorithmic governance is only as good as the human inputs for and through which it is constructed, operated, supervised, and valued. But perhaps as important is the insight that international organizations themselves may require reform to reduce the potential for systemic corruption that may be inherent in the ways in which these organizations are constituted, staffed, and operated. Corruption synergies existed across the structures of the World Bank and from within it to the way in which it related to key stakeholders respecting the production of objects (ratings) of intense interest to these stakeholders. Yet it is precisely the inherent structural element of inter-institutional corruption that remains untouched. For now. The stakes are as high in the context of data driven governance as they have been under the prior regimes of law and bureaucratic administration systems.
 |
| Corruption of Video Feed; Pix Credit HERE |
And yet, a careful reading of the Report also suggests the difficulty of
disciplinary measures in under conditions of power politics. Like law,
data governance will likley suffer from the realities that it can do
little more than expresses the desires of those with the power to control
access to data, to mold the principles under which analytics are
validated, to deploy the analytics strategically, and to develop
algorithms and the applications of judgments drawn from them in interest
enhancing strategic ways. If the system is law is structurally
compromised (in its search for perfection) by the inherent flaws of the
human for whose welfare law is deployed, systems of quantitative,
predictive and simulated mechanisms developed for those same ends will
likely suffer the same impediments to perfection. But those corruptions
will (until the language of these systems is mastered) be more easy to
hide behind ideologies that presume the omnipotence, neutrality, and
fixity of quantitative measures.
The World Bank Statements and the WilmerHale Report follow.
