From inclusion to innovation: reimagining AI development through diverse knowledge systems

This paper focuses on inclusion, questioning the status quo of knowledge production and emphasising the potential for diverse contributions to AI development. When AI collects data about humans, it acts as a mirror, reflecting the stereotypes and inequalities present in society. Scientists and scholars argue that these biases often arise from a lack of diversity among those developing AI systems and tools, as well as from biases embedded within science and culture. AI teams and developers often have gaps related to their technical training, areas of expertise, and cultural and social backgrounds, which are inevitably reflected in the tools they create. This paper argues that the voices and perspectives of people who speak different social languages and who think and act in ways that diverge from the status quo must be included. It presents a case study based on experiences from an EU-funded project. This approach moves AI development beyond the current cycle of knowledge production, enabling new developers and professionals to adapt tools to the needs and experiences of their diverse communities through collaboration across disciplines, cultures, scientific fields, and related approaches. The thesis is that fostering a deeper and more substantive dialogue between the social sciences – particularly sociology, education, and psychology – and STEM is essential for developing a science that is inclusive from the outset and at the design stage.