Transduções informacionais e Machine Learning: reflexões sobre possível inciência de atores envueltos

The study aims to discuss aspects of the unawareness of the actors involved in the development and use of Machine Learning models, focusing on the informational transductions that occur throughout their stages, that is, the transformations that information undergoes across the different phases of the data and model life cycle, from collection to application. The methodological procedure adopted is based on descriptive, exploratory, and qualitative research, and data were collected from the Hugging Face and GitHub platforms in order to analyze the possible transformations and informational transductions that occur during the training phase of the available models. The results indicate that, although these platforms offer a large volume of models and tools, there is opacity regarding the transformations applied to the original data, which limits users’ understanding of the integrity and reliability of the available models. It is therefore concluded that there is a need for platforms to adopt transparency protocols and standardized descriptive structures to represent the transductions and technical decisions involved.