Why you should care: Ethical AI principles in a business setting : A study investigating the relevancy of the Ethical framework for AI in the context of the IT and telecom industry in Sweden

Abstract: Background: The development of artificial intelligence (AI) is ever increasing, especially in the telecom and IT industry due to its great potential competitive advantage. However, AI is implemented at a fast phase in society with insufficient consideration for the ethical implications. Luckily, different initiatives and organizations are now launching ethical principles to prevent possible negative effects stemming from AI usage. One example is the Ethical Framework for AI by Floridi et al., (2018) who established five ethical principles for sustainable AI with inspiration from bioethics. Moreover, Sweden as a country is taking AI ethics seriously since the government is on a mission to be the world leader in harnessing artificial intelligence. Problem: The research in the field of ethical artificial intelligence is increasing but is still in its infancy where the majority of the academic articles are conceptual papers. Moreover, the few frameworks that exist for responsible AI are not always action-guiding and applicable to all AI applications and contexts. Purpose: This study aims to contribute with empirical evidence within the topic of artificial intelligence ethics and investigate the relevancy of an existing framework, namely the Ethical Framework for AI by Floridi et al., (2018), in the IT and telecom industry in Sweden. Method: A qualitative multiple-case study of ten semi-structured interviews with participants from the companies EVRY and Ericsson. The findings have later been connected to the literature within the field of artificial intelligence and ethics. Results: The most reasonable interpretation from the findings and analysis is that some parts of the framework are relevant, while others are not. Specifically, the principles of autonomy and non- maleficence seem to be applicable, meanwhile justice and explicability appear to only be partially supported by the participants and beneficence is suggested to not be relevant due to several reasons.