The Bernoulli Principle: Epidemiological Models for Identity
Smallpox is one of the deadliest diseases in all of human history. If you contracted Smallpox in the 18th century, you would have a one in three chance of dying within sixteen days. In the 20th century alone, Smallpox killed over 500 million people. Fortunately, it’s also the only infectious disease among humans that has been successfully eradicated.
Daniel Bernoulli, attempting prove the efficacy of inoculation against Smallpox, published the first epidemiological model in 1760; he demonstrated that life expectancy increased due to the use of inoculation against Smallpox in the general population. In doing so, he introduced the use of epidemiological models that have been used to address the spread of not only Smallpox, but also malaria, AIDS, SARS, measles, cholera, etc.
Bernoulli’s model introduced provided three separate benefits related to the spread of disease: an understanding of the mechanism of transmission, a prediction of the future expansion of infection, and, of course, control over the spread of the disease. A machine-trained model based on this epidemiological model, when combined with a network-graph representation of identity, has the potential to provide similar results related to the spread of identity and its related access. In short, applying epidemiological concepts to identity holds great promise for innovation.
Just as smallpox was communicated from patient to patient, access to sensitive data and applications often spreads like a disease inside communities of identities. By understanding how access is communicated from “patient zeros” to the surrounding community, it is possible to begin to predict which identities will soon accumulate access, and then seek to discover inoculation-type tactics to restrict the spread of unnecessary access. By analyzing these “infection patterns,” the machine-trained model can provide recommendations for governing identity, enhancing decision-making, educating human users, and pairing machine learning and human learning in a “virtuous loop.” Over time, routine approvals or revocation of access could be completely automated, allowing humans to focus solely on difficult boundary cases, accelerating overall productivity for securing identity. By thus “inoculating” communities against the rampant spread of access, risk to enterprises and the community at large is reduced.
There are times, however, when a model of this type should be used to promote the spread of identity rather than to restrict it. Initiatives such as ID2020 are endeavoring to ensure that underserved groups are not left behind by the promise of digital transformation: by granting them identities, they clear a path for them to access health care, exercise their voting rights, obtain education, or otherwise reap the benefits of what are assumed as basic human rights. This model could be used to examine how underprivileged communities adopt identity, seek to remove inhibitors to its acceptance, and accelerate its adoption in communities worldwide.
Thus, what Bernoulli began in 1760 still finds its expression today: modeling the real world with the end goal of reducing harm and improving the quality of human life. As we seek to innovate in his footsteps, we have no doubt that he would be using the same techniques and ideas were he around today.