Taking Responsibility for Diversity & Bias in AI

AI is being presented as a neutral agent but biases can be detected and many conclude that this is a result of the lack of diversity of actors in the development of AI. AI can be a force for good but if it reflects and amplifies existing inequalities this risks compounding the impact of these inequalities and providing information which is at best unfair and at worst ineffective or damaging to groups of individuals, for example if it replicates existing biases in medical diagnosis of certain diseases and conditions in ethnic minorities, or the production of misogynist materials. Finding ways to bring more diversity into the workforce is critical so addressing barriers into training and employment are critical. To be effective at addressing bias and lack of diversity, it is essential to address questions of responsibility and the implications of this. Whilst the AI industry can and arguably should take responsibility for the content and impact of the output, a biased system cannot adequately achieve this task without bringing in other voices and expertise. This event provides a forum for examining these principles and the operating practices which can follow, bringing people with expertise in equalities into the discussion with AI specialists and educationalists, to shed light on how bias can be prevented as well as identified and how the eco-system of training, employment and data mining can be a better force for good by addressing bias and lack of diversity.