THE DANGER OF LEGITIMACY LAUNDERING

Biased indexes can accrue credibility in an echo chamber of cross-references. Reforms in the World Bank's governance indicators show a structured pushback can yield better results

- SANJEEV SANYAL

Narrative wars have been around from the beginning of civilization. Kingdoms, republics, tribes and rebel groups have long used disinformation and ideology to manipulate political narratives to their advantage. The spread of digital networks and social media, however, has allowed for an unprecedentedly rapid penetration of such narratives. Artificial intelligence now presents the danger of an algorithm-driven mass-scale manipulation that all sovereign governments and citizens need to guard against.

One could argue that people will adjust their behavior to the new environment and become more skeptical. Notice, however, that sources that establish themselves as credible become very important in such an environment of doubt. Therefore, the new battlefield is for the 'legitimacy' and 'credibility' of information sources. The labeling of a source as credible is especially powerful as its data would be used to train AI platforms that then run the default responses of many systems.

Those attempting to manipulate global narratives are aware of the credibility issue and have been investing in taking control of default sources even as they have been taking great pain to mask the takeover. One tactic, drawing inspiration from academic publishing, is to create cross-references in order to echo a particular view so that it comes to be seen as the accepted truth. Akin to money laundering, legitimacy laundering then creates layers of sources that hide the original manipulators.

Notice how, in recent years, there has been a proliferation of global indexes and rankings. These rankings are themselves derived from other indices-often entirely subjective opinions-and, in turn, flow into sovereign ratings, investment weightages and so on. In this way, real-world decisions get influenced by sources that are not easily visible to the end user.