New technologies and mixed-use convergence: How humans and algorithms are adapting to each other

Abstract

Human experience with technology has shifted from technological contexts requiring occasional intervention by a fraction of people mostly in command of technologies, to technological contexts that require constant ongoing participation from most people to complete tasks. We examine the current state of `mixed-use' new technologies integration with legacy systems, and whether the human assistance required to complete tasks and processes could function as a training ground for future smart systems, or whether increasing `co-dependence with' or `training of' algorithmic systems, enhancing task completion and inadvertently educating systems in human behaviour and intelligence, will simply subsume people into the algorithmic landscape. As the Internet of Things (IoT) arises in conjunction with advancing robotics and drone technology, semi and fully automated algorithmic systems are being developed that intersect with human experience in new and heterogeneous ways. Many new technologies are not yet flexible enough to support the choices people require in their daily lives, due to limitations in the algorithmic `logics' used that restrict options to predetermined pathways conceived of by programmers. This greatly limits human agency, and presently the potential to overcome problems that arise in processes. In this mixed-use period, we have the opportunity to develop new ways to address ethical guidance as knowledge that machines can learn. We explore promoting embedding of ethically-based principles into automated contexts through: (1) developing mutually agreed automated external ethical review systems (human or otherwise) that evaluate conformance across multiple ethical codes and provide feedback to designers, agents, and users on the distribution of conformance; (2) focusing on review systems to drive distributed development of embedded ethical principles in individual services by responding to this feedback to develop ongoing correction through dynamic adaption or incremental releases; and (3) using multi-agent simulation tools to forecast scenarios in real time.