Text and Multimodal Data Processing for Sustainable Development

It is now being trotted out to define our existence with LLM and AI, signifying an era in which Large Language Models have seamlessly elbowed their way into practically every environment containing computing power and an internet connection. It is a preoccupying use of the prefix, not only because of its marketing-like potential to be self-prophetic, but also because of how derivative it makes our reality sound. Framed within a much bigger picture, this is worrying at a time when basic human rights are being challenged the world over. Despite the challenges we face with the massive adoption of LLM, they do provide us with a formidable heuristic with which to study how humans function linguistically. By studying the human linguistic response in hybrid productions (post-edit) the Human Inside project is aiming to reveal the elements of human language that are ‘robust’ enough in their representation to be deemed worthy of inclusion in the final version. Language is dynamic and ever-evolving, and words fade and new ones appear all the time, sometimes naturally, sometimes via authoritarianism. One major problem of having one’s lexicon redefined by successive sweeps of algorithmic interaction is that even though the human at the keyboard is sentient and independent, the sheer power and stickiness of suggestion means that only the best educated and/or the strongest willed can complete a hybrid writing process with their initial idea, intention, and organic lexicon fully intact. And since the ‘imperfection’ of human language often contains semantic material and not just stylistic or functional chaff, these practices must be addressed.
The Human Inside project is an international, multilingual project bringing together language science researchers working on hybrid LLM productions in Mandarin, French, Vietnamese, Portuguese, and English, which allow us to have long-term goals that include cross linguistic analyses of commonly identified features. During this talk I’ll present an overview of how the Human Inside project functions, from experiment design to data collection, object identification, corpus construction, and statistical analysis. We’ll then cover a case study from my own particular focus on this project, which is embodied communication, and its potential to represent offline human ecology within text. It is important for our heritage in the making that we identify the words that link us to our ecologies and environments, because the risk is—to use an agricultural analogy—that we end up with semantic monocultures which will negatively impact our language and communicative abilities. This is all the more of an issue for places of emergent economy in the global South that risk passing from heavily oral traditions to computer-mediated communications without the intermediary accumulation of literature, documentation, semiotic circulations, and writing praxes that much of the western world was able to accrue prior to AI. Finally, we’ll see how this might be achieved by ‘digital rewilding’ in order to reinstate the semantics of human idiosyncrasy and ecology in written productions as a feature, not a bug. This would be achieved via application, not static archiving, because unlike plant seeds, words live in circulation, and die in archives.

In the business world, AI enhances strategies, increases productivity, and creates opportunities for innovation and growth. In the field of education, it enables personalized learning experiences and equips students to face real-world challenges with confidence. His keynote will highlight how the synergy between technological innovation and human vision can drive meaningful impact and lasting transformation.

We will detail methods for crafting multimodal narratives that leverage visual, textual, and interactive elements to communicate scientific insights clearly and persuasively. A central focus will be on the application of these strategies specifically toward the SDGs, examining how targeted data communication can effectively frame problems, demonstrate progress, and mobilize support.

Furthermore, we will discuss foundational aspects of Scientific Data Management, emphasizing the importance of high-quality, trustworthy data. This includes an in-depth review of the FAIR Data Principles (Findable, Accessible, Interoperable, and Reusable), showing how adherence to these standards is essential for building transparent, reproducible, and impactful data narratives that underpin global sustainable development efforts.