Minorities exclusion, a shame for humanity in the age of AI.

Published On: November 2021Categories: 2021 Conference Editorials, Editorials


Christian E. Maldonado-Sifuentes

Centro de Investigación en Computación del Instituto Politécnico Nacional México

PhD Candidate

Photo of a bald man in a red polo shirt

In a world where Artificial Intelligence has advanced so much, so fast, no single human should be excluded for not speaking a dominant language. As of today most of the advances on AI language research has been done for a handful of dominant languages, excluding —almost completely— the thousands of minority and endangered languages of the world, making the digital divide even deeper at ever accelerating rates.


This is not a tirade against the insensitivity and aloofness of the research or governmental elites. It is a serious call for action, since research in Artificial Intelligence will not gravitate naturally towards the inclusion of minorities. There are many reasons for this, and I will try to address them succinctly but comprehensively in this piece.


Firstly, the biggest technological hurdle: Modern AI techniques tend to work better with —almost beyond comprehension— massive datasets. The great advances that have been attained in the last 20 years of AI research have come, in a great deal, from the advent of the internet and its massive amounts of data. This data has in turn allowed more complex algorithms to learn subtle language patterns. The lack of (or late) access to this emerging media for language minorities, either through inequality or disability, has prevented the same volumes of information to be produced for these languages. There are many possible solutions, but scientific ingenuity could play a big role in finding ways to make modern techniques work with smaller data. This is both an interesting research problem and necessary to bridge the gap between dominant and minority languages.


Secondly, the incentives for the scientific community and even their viability depend on evaluation criteria that discourages research for good causes. In the field of computer science, but especially in AI research most of the works try to “beat the performance of the state-of-the-art”. While this affects the way AI research is done in general, is specially nocive for research that could have social impact but will not have immediate fruition. Most researchers, research groups and even institutions look to be competitive in their area rather than risk losing their funding. Creativity is needed from policymakers to devise methods that preserve competitiveness and productivity but also reduce the risk researchers have to make while trying to tackle these social problems.

Finally, there is a lot of work that has already been advanced by social sciences and humanities with respect to the inclusion of linguistic minorities. These disciplines are allowed to —and perhaps even rewarded for— presenting works with a social impact. Still, most of these studies and developments are done with technologically outdated techniques. Likewise, most of the work tends to be disperse in nature with little chances of producing the easily consumable and vast data that AI researchers yearn for. Both scientists and policymakers are needed to bring out their best and promote ways for collaboration between the AI community and the social disciplines community. Tools that favor the work of social researchers can be developed by the technological side of the aisle, this in turn, could facilitate the creation of the large repositories that would enable AI minority languages research, applications and innovations. These efforts ought to be promoted by sound and decisive policy that empowers scientists to develop transdisciplinary work.

We already have some examples from other areas where technological development is fostered to some extent, like renewable energies or smart cities. While these challenges are more urgent because of the impact they could have on the protection of the environment and the economy, the social and economic impact of inclusion is not negligible by any means. A completely connected society where the barrier language is a thing of the past, would create a cascading effect that would diminish exclusion, discrimination, ostracism and alienation of minorities. This, in turn expands the universe of workforce and consumers and has the potential of boosting the economy as well.

The inclusion of minorities is not only a moral imperative, but a benefit to the society as a whole.