Paraguayan student Samantha Adorno has presented an innovative technological proposal at an international conference in Barcelona, outlining a new approach for developing artificial intelligence capable of understanding oral languages such as Guaraní. Samantha, a third-year computer science student at the University of Kansas, detailed her AI to integrate Guaraní model at the prestigious CHI 2026 congress, aiming to transform the relationship between technology and language.
Her work stems from a key premise: artificial intelligence should not be limited to processing text, but must learn to listen and comprehend the cultural context of its users. This is particularly crucial for languages with strong oral traditions, which are often left behind in a text-dominated digital world. Samantha’s presentation represents a significant step towards creating more inclusive and accessible technological solutions for diverse linguistic communities.
AI to integrate Guaraní, bridging the digital divide
During her address, the young researcher highlighted a significant disparity in her home country. Although Paraguay recognises both Spanish and Guaraní as official languages, the latter remains profoundly underrepresented in digital environments. This is despite Guaraní being spoken by a large portion of the population. She argued that this gap is largely due to the language’s history as a predominantly oral form of communication.
This oral nature makes it difficult to adapt to technological systems that have been designed primarily for written text. Consequently, existing AI models and digital tools often fail to serve Guaraní speakers effectively. Samantha’s work directly confronts this challenge by rethinking the fundamental architecture of language-based AI to better accommodate the nuances of spoken communication and cultural expression.
A new architecture for inclusive AI
In response to this issue, the student proposed an artificial intelligence architecture built upon multiple specialised agents. This system is designed to process voice, interpret user intentions, manage conversational context, and crucially, guarantee consent for the use of data. Such a multi-faceted approach moves beyond simple transcription, and aims for a deeper level of comprehension.
This innovative framework would allow for the development of systems that more closely mirror the real communicative practices of communities. Therefore, it avoids the common problem where users must unnaturally adapt their way of speaking to fit the limitations of technology. The core objective, as Samantha explained, is to ensure that technology adapts to people, not the other way around, thereby respecting not only the language but also its rich cultural and social context.
Emphasising data sovereignty
A central pillar of the proposal is the concept of data sovereignty. Her model insists that linguistic communities must have complete control over how their information is collected, used, and managed. This principle is vital for protecting the integrity of indigenous languages and ensuring that technological advancements benefit the communities they are intended to serve.
This focus on community control marks a departure from traditional data collection models, which can often be extractive. By empowering communities to govern their own linguistic data, the project fosters a more ethical and collaborative approach to AI development. It ensures, according to the student, that cultural heritage is preserved and respected throughout the technological process, making communities active partners in innovation.
Future plans and international recognition
The AI to integrate Guaraní work presented by Samantha was praised as a significant contribution to the field of human-computer interaction. It opens new possibilities for the inclusion of indigenous and oral languages in global technological development. Furthermore, the researcher announced her intention to continue the project in Paraguay, where she plans to incorporate the direct participation of local communities to strengthen and refine the model’s design.
