The Grammar of Who Gets to Know

In Decolonising the Mind (1986), Ngũgĩ wa Thiong’o describes a school in Kenya where a pupil caught speaking Gikuyu was punished or made to wear a metal plate around the neck with inscriptions such as ‘I am stupid’. The same book was his farewell to English as a language of writing. A few years earlier, in Kamiti prison, he had written his first novel in Gikuyu on toilet paper. We usually read this story as a story about colonialism. I want to read it as a story about knowledge: the lesson of that metal plate was which language had the right to carry knowing.

My question today is whether that plate has really been lifted from our necks or has only changed its shape. Today’s plate is not made of metal. It is an index that cannot see an article in Kiswahili, a funding application that can only be filled in in English, and a language model that speaks fluently about farming on the slopes of Mount Kenya because other people wrote about it in English, not the farmers who know those slopes.

From the hearer to the archive

Miranda Fricker (2007) gave us two concepts without which it is now impossible to think about knowledge. Testimonial injustice occurs when a hearer, out of prejudice, gives a speaker less credibility than they are due. Hermeneutical injustice occurs when a group’s experience finds no words at all in the shared resources of understanding. Kristie Dotson (2011) showed that this silencing has other forms too: a speaker who knows they will not be heard shortens what they say in advance, or does not say it at all.

In both formulations the injustice is located in a relation between two people: one speaks and the other hears. But most knowledge in circulation today does not pass through a human hearer. It passes through citation databases, library catalogues, search engines and the training data of models. My proposal is that we treat epistemic injustice as a property of these infrastructures as well, not only of individual judgement. Prejudice can be built into the design of a database without anyone exercising it at the moment it takes effect.

Languages of record and languages of report

I distinguish between two kinds of language. A language of record is one in which knowledge is indexed, cited, counted and credited towards a career. A language of report is one in which knowledge is spoken but not counted. A farmer who explains in Gikuyu how rain and soil move through the year becomes an ‘informant’ in the article a researcher publishes in English, not an author. Her knowledge travels from the language of report to the language of record, and her name is left behind on the way.

Boaventura de Sousa Santos (2014) speaks of ‘epistemicide’: the destruction of ways of knowing together with the peoples who carried them. What we face now is quieter, and for that reason more dangerous. No knowledge is killed; it simply remains in a language the indexes do not read. When large language models have learned mainly from English text, they reproduce this gap at an unprecedented scale: the language of record grows more fluent by the day, and the language of report more silent.

What has to be built

If injustice sits in infrastructure, justice has to be built into infrastructure too, not left to the personal ethics of researchers. Three things seem to me essential. First, authorship in the language of origin: whoever said the knowledge in Gikuyu or Kiswahili is its author, and that name must stay with the text in every translation. Second, translation as representation, not replacement: the English text must not hide the original but always lead back to it. Third, counting anew: as long as only what is published in languages of record is counted, no multilingualism policy can be taken seriously.

Ngũgĩ paid a price for writing in Gikuyu. We should not ask young researchers to pay that price again. Our work over these three days is to ask what an infrastructure that holds knowing hostage to no language would look like, and who should govern it.

Questions for discussion

References

Ngũgĩ wa Thiong’o. 1986. Decolonising the Mind: The Politics of Language in African Literature. London: James Currey.

Fricker, Miranda. 2007. Epistemic Injustice: Power and the Ethics of Knowing. Oxford: Oxford University Press.

Dotson, Kristie. 2011. “Tracking Epistemic Violence, Tracking Practices of Silencing.” Hypatia 26 (2): 236–257.

Santos, Boaventura de Sousa. 2014. Epistemologies of the South: Justice Against Epistemicide. Boulder: Paradigm Publishers.

Spivak, Gayatri Chakravorty. 1988. “Can the Subaltern Speak?” In Marxism and the Interpretation of Culture, edited by Cary Nelson and Lawrence Grossberg, 271–313. Urbana: University of Illinois Press.