Building Artificial Intelligence for Digital Health: a socio-tech-med approach and a few surveillance nightmares

In this paper, I focus on two main goals: first, I try to empirically deconstruct algorithms opacity by opening up their «black box» through my participation in an interdisciplinary group of experts. We designed an artificial intelligence project for the early prevention of heart stroke (AICARDIO). Secondly, I demonstrate to a non-expert audience how our (big) data feed artificial intelligence algorithms, easily and unnoticeably, in our everyday lives. I reversed my role from being an observed «subject» of security surveillance to becoming a «designer» of algorithm technology for patients’ health surveillance, struggling to protect data privacy. I analyze several empirical examples (the meta-observer; without data, algorithms are useless; algorithmic literacy; the project’s soul; face emotion recognition; and privacy matters). Algorithms do not appear as SW-engineered instructions to carry out tasks. My research will show that the mediating processes of our interdisciplinary group of experts lead us to the construction of algorithms bridging the tension and mediation of multiple socio-tech-med cultures and human bias. The invisibility of AI-based surveillance technology seems to be a controversial issue in many domains, for example, home-based health monitoring systems. A heterogeneous ecosystem competes in complex human and non-human practices in algorithmic societies, creating an autopoietic ecosystem affecting our everyday lives, that still needs a rapidly evolving and human-centric AI governance.