Researchers’ Concerns on Artificial Intelligence Ethics : Results from a Scenario-Based Survey

Abstract

The ethical impacts of Artificial Intelligence (AI) are causing concern in many areas of AI research and development. The implementation of AI ethics is still, in many ways, a work in progress, but various initiatives are tackling the issues by creating guidelines and implementation methods. This study investigates concerns about the negative impacts of AI systems posed by researchers working with AI. The study was conducted as a scenario-based survey, in which participants answered the question, "What could go wrong?" regarding five scenarios depicting fictional AI systems. The study concludes with the results from 33 survey participants who gave 161 responses to the scenarios. The results suggest that researchers can identify threats posed by AI systems, particularly regarding their social and ethical consequences. This is even though half of the participants reported limited involvement with AI ethics in their work. The widespread understanding of ethics among researchers could positively impact AI software development due to increased capabilities to bring theoretical AI ethics to practice.