Social Innovation on Educational AI Developments. A Case Study on Social Participation on Designing AI Generative Models for Diversity

Social engagement processes are critical to technology development if we want to ensure inclusivity, equity, and responsiveness to diverse populations. Generative AI tools require new models of technology development due to the complexity of their structure, from the creation of large datasets to the adjustment of biases to ensuring fairness of their results. The use of community-based research approaches to involve diverse groups and citizen sensitivities in the research and development process seems a good strategy to guarantee the ethical design and use of these knowledge-generating technologies. A particular case it is presented regarding the potential of AI generative tools to simplify legal, administrative, or educational texts by providing new translation tools. This is a multidisciplinary project that requires consideration of ethical and legal consequences and that can be addressed by a social engagement development model. Requires civil and democratic engagement in two areas: 1) democratically selecting the type of texts and translations needed, and 2) involving citizens, experts, and non-experts to produce and validate real examples of legal texts with cognitive adaptations. These adaptations will feed the AI algorithms used to learn how to translate the texts. The social value of this model extends beyond the design of a text simplification tool. The model aims to establish participation and validation criteria to prevent legal, social, and equity biases in the design of generative AI tools for social purposes, improving the design of validation and explainability processes guided by human experts.