Illustration by Somnath Bhatt

The Power and Politics in Translation

A guest post by Asvatha Babu. Asatha is a PhD candidate at American University’s School of Communication and a Doctoral Researcher at the Internet Governance Lab, studying emerging technologies and new media. Twitter: @fireflyclass

This essay is part of our ongoing “AI Lexicon” project, a call for contributions to generate alternate narratives, positionalities, and understandings to the better known and widely circulated ways of talking about AI.

There is no word for facial recognition technology (FRT) in Tamil. Instead, local news reporters talking about the technology in Tamil Nadu, India, have resorted to the use of close alternatives like mugam aridhal thozhilnutpam (face knowing technology) or mugathaal adaiyaalam kaattum thozhilnutpam (technology to identify through face). The Tamil words for system (amaippu) or app (seyali) are sometimes used instead of thozhilnutpam (technology). So already we have at least six different phrases that all refer to the same technology. Add to this the multiple different ways that these words could be grammatically combined, and we get a rather long list of names for a single technology.

For this essay and as a part of my doctoral dissertation, I analyzed Tamil-language news coverage about FRT in India to understand: the complexities of translating technological concepts into another cultural context; the key stakeholders translating technologies to local contexts; and how this translation shapes the local construction of FRT. I argue that the complex ways in which FRT gets translated are shaped by the politics of the actors who are primarily charged with doing the translation work. And in turn, these translations shape the local cultural construction of the technology in ways that are in line with those politics.

According to cultural historian Leo Marx (2010), the emergence of a singular name or keyword often serves as a marker for cultural change. Without that keyword, a semantic void arises that makes it harder to describe related societal changes. (Marx, 2004 as cited in Oldenziel, 2006). In the case of FRT in Tamil, journalists have attempted to fill the void by directly transliterating the phrase “facial recognition” (WebDesk, 2019). In so doing, an opportunity to take local ownership of the technology is lost. Facial recognition systems and the software underpinning them are often already proprietary black boxes. When they continue to be referred to with an English name, they become perceived as foreign objects disconnected from the local context, another blackbox that is not constructed or shaped locally and hence inaccessible to those interested in participating in its construction.

The complexities of translating technological concepts into multiple languages and cultural contexts throw sharp light onto the politics that go into the cultural construction of technology. Take algorithmic fairness for instance. Sambasivan, Arnesen, Hutchinson, and Prabhakaran (2020) show how difficult it is to technologically apply the concept of fairness in the same way in India as in the US due to differing contexts, histories, and social structures. The word “fairness” itself can also be constructed differently in different languages. In Tamil, for example, fairness can be translated to either nyaayam (justice) or nadu-nilai (neutral). In order to translate fairness into Tamil, one needs to first answer what is truly meant by algorithmic fairness: algorithmic neutrality or algorithmic justice? And who decides this?

The lack of a specific Tamil name for “facial recognition” has clearly shaped how the technology is locally constructed, used, and perceived. Law enforcement organizations in Chennai, India have been using locally developed FRT over the past three years in public surveillance, as a technological aide during daily patrols to identify those they find suspicious, and in other policing practices (Pawar, 2018). It is also used by some private organizations to regulate access (Hariharan, 2020), and some schools to mark attendance (Pon Vasanth, 2019). Without a buzzword-y name marking it as a discrete object, Tamil media coverage of this use tends to foreground the purposes of a particular application of FRT (to find missing children or identify criminals, for example) rather than question the technology itself (Na Pa Sethuraman, 2018; Dhinamalar, 2019). In other words, articles are more likely to be headlined that a particular problem is being addressed by technology generally, rather than discussing or addressing the concerns of facial recognition on its own. Facial recognition thus does not get treated in the local discourse as a discrete entity worthy of critique on its own.

My analysis of media coverage also shows three key voices who translate and thus dictate the conversation around FRT and related concepts in Tamil: 1) the local engineering and Tamil scholarly community; 2) the technologists developing FR locally; and 3) Google Translate. Among those missing are social scientists, civil society, human rights organizations, and policymakers.

Engineering and Tamil Scholarly Community: The translation of Facial Recognition into the Tamil phrase mugam aridhal can be traced at least to 1998, when one of the foremost engineering universities in Tamil Nadu — Anna University — released their first lexicon of computing terms translated into Tamil. Based on American mathematician Donald C. Spencer’s The Illustrated Computer Dictionary, the first lexicon has over 4000 computing terms translated into Tamil by a council of 9 engineers and 3 Tamil scholars. These experts developed this deeply researched lexicon in an effort to grow the Tamil language so that it covers the technical field and to make engineering education accessible to more Tamil people. For example, they translated the term algorithm as Nerimurai, where neri means norm and murai means “the way to do something.” Together they can mean a set of rules to do something. Nerimurai is also used to refer to a generally accepted protocol for doing a particular task. They also translated artificial intelligence (seyarkkai nunnarivu), and various kinds of recognition (all to ___ aridhal). Although facial recognition technology was curiously absent, it is easy to deduce that it should be translated as mugam aridhal (Face knowing) based on the other translations in the list (S. Srinivasan, V. Krishnamoorthy, A. Ilangovan et al, 2010).

FRT drivers: The drivers of FRT in this case include the technologists developing the tech in Chennai as well as police officials who drive this development and comprise the primary user base. They have played a role in shaping the discourse of the technology through their interactions with the media and their communications with the public on social media. While these drivers largely use the Tamil transliteration of “facial recognition,” their description of the technology using certain words in certain contexts plays a role in framing the technology for the larger public (BBC News, 2019). For instance, a recent article about the use of FRT enabled COVID-19 surveillance app CoBuddy was described by the developer and police officer in charge as “a friend when someone is in need of COVID related assistance” (S Magesh, 2020). The translation practices of these technologists served to legitimize FRT as a familiar, friendly, technological aid to a more convenient and safer life. Researchers, on the other hand, have described CoBuddy app more fully as a geo-fencing app to monitor “the movement of COVID-19 suspects under home quarantine” (Bajpai and Wadhwa, 2020).

Google Translate: Google has also played a surprising, yet indicative, role in (mis)translating facial recognition into Tamil. Although the translation of facial recognition into “muga angeegaaram” by Google Translate does not entirely make sense in Tamil, it has been adopted by news articles (WebDesk, 2020) and is the title of facial recognition’s Tamil Wikipedia page. Shedding all context (as Google Translate often does), angeegaaram indeed literally translates to recognition. However, recognition in this case refers to the formal or official acknowledgement of the existence of an entity (a non-profit organization being “recognized” as such by the government, for example). The uncritical and continued adoption of this term by Tamil journalists is significant because the word angeegaaram in Tamil comes loaded with ideas of government approval and authorization. This (inaccurate) association of government approval with FRT can play a role in shaping how the public thinks about this technology even when it is deployed by private players with questionable motives. It also indicates that journalists may have placed a higher valence on Google’s translation accuracy than their own which is worrying for the development of a more robust discourse on technological concepts such as fairness, surveillance, and privacy.

To translate technological ideas from English to other languages is to make a series of complex decisions about the various aspects of the technology. It is no simple task and benefits from the involvement of a variety of stakeholders. The engineering and Tamil scholarly community’s joint engagement to create new ways of talking about technical ideas is significant. In addition to growing their respective fields, by putting out a jointly compiled official lexicon of translated terms, they have offered a model for a decision-making process that lightens the weight of any one stakeholder decision, which may or may not consider the context or complexities of the technology. More such conscious engagements could also foster a rich discourse about important related ideas such as risks, harms, power relations, privacy, surveillance, fairness, and others. Further, the involvement of people from policy and civil society communities could help bring in works of other Indian scholars working on technology rights in other languages which can also enrich the translation process.


Bajpai, N., & Wadhwa, M. (2020). Artificial Intelligence and Healthcare in India (#43; ICT India Working Paper). Center for Sustainable Development.

Hariharan, S. (2020, October 17). Facial recognition technology, AI drive back-to-work rules. Times of India.

Marx, L. (2000). The machine in the garden: Technology and the pastoral ideal in America. Oxford University Press, USA.

Marx, L. (2010). Technology: The emergence of a hazardous concept. Technology and Culture, 51(3), 561–577.

Na Pa Sethuraman. (2018, April 18). குற்றவாளிகளை அடையாளம் காட்டும் ஆப். Vikatan (Online).

Oldenziel, R. (2006). Introduction: Signifying semantics for a history of technology. Technology and Culture, 47(3), 477–485.

Pawar, J. (2018, July 4). Facetagr app: Chennai police’s bright spark helps nab elusive criminals. The New Indian Express.

Pon Vasanth. (2019, September 14). Face recognition attendance system in two Chennai schools. The Hindu.

S Magesh. (2020, April 4). `மூச்சுவிட முடியல.. தொண்டை வலிக்கிறது!’ — தனிமைப்படுத்தப்பட்டவருக்கு Cobuddy ஆப் மூலம் உதவிய எஸ்.பி.

S. Srinivasan, V. Krishnamoorthy, A. Ilangovan, Ulo. Senthamizh Kodhai, P.A. Nakkeeran, R.T. Rasan, M. Aarumugam, D. Che. Subburaman, A. Mathiyazhagan, P. Ramalingam, R. Ilavarasu, & Kodumudi S. Shanmugan. (1998). கணிப்பொறி கலைச்சொல் அகராதி [Lexicon]. Anna University.

Sambasivan, N., Arnesen, E., Hutchinson, B., & Prabhakaran, V. (2020). Non-portability of Algorithmic Fairness in India. ArXiv Preprint ArXiv:2012.03659.

WebDesk. (2019, July 10). Facial Recognition: குற்றவாளிகளுக்கு ‘செக்’ வைக்கும் இன்னொரு புதிய வசதி. The Indian Express (Tamil).

WebDesk. (2020, June 12). முக அங்கீகார தொழில்நுட்பத்தை போலீஸ் பயன்படுத்த அமேசான் தடை. Puthiya Thalaimurai.

காணாமல் போன குழந்தைகளை கண்டுபிடிக்கும் தொழில்நுட்பம். (2019, June 2). BBC News.

காணாமல் போன குழந்தைகளை கண்டுப்பிடிக்க புதிய ஆப் அறிமுகம். (2019, June 30). தினமலர் (Dinamalar).