AI risks: An organisational practice approach to trustworthiness

Paper in proceeding, 2024

Artificial intelligence (AI) is in need for a framework that balances the opportunities it represents with its risks. But while there is a broad consensus on this, and public regulative initiatives are taken; there is far less knowledge about how these dilemmas/opportunities/risks are played out in practice. The interest into ethics in organisation driven by a discourse on "Trustworthy AI"; makes us wonder whether an ethical approach to AI in organisation is purposeful; or needs modification. We investigate this by viewing the development and use of AI as structuration of practices. The empirical material is our own development of an AI system. Using studies of ethics in moral engineering design; AI is a question of structuration processers with unintended consequences. It is a "slide" from ethics of virtue to ethics of benefit as corroborated by engineers/designers referring ethical dilemmas to managers and politicians. The EU framework of Trustworthy AI for designing and using more accountable AI systems-considering ethics; human autonomy; harm prevention; fairness etc., conflicts with contemporary construction organisations. We propose an extension of the EU guidelines.