Images of AI: How AI practitioners view the impact of Artificial Intelligence on society, now and in the future

Despite unprecedented technological advancement, intense commercial investment, international agreements, and growing societal concerns with Artificial Intelligence (AI), there is little insight into how those driving the field – the everyday AI practitioners – perceive AI and its impact on society, now and in the future. We address this critical gap by conducting a broad-based survey with 100 AI practitioners, followed by in-depth interviews with 20 AI practitioners, including developers, managers, and consultants. Using socio-technical grounded theory (STGT) for data analysis, we inductively identified six images of AI which capture six ways in which AI practitioners view AI, now and in the future, and their implications for impact on society and human control : Parrot captures AI that mimics human behaviour, including biases; Companion surrounds humans in daily life and supports decision making with empathy-like traits; Wolf in Sheep’s Clothing highlights AI misused by humans, causing societal harms; Saviour envisions AI solving complex problems beyond human capacity; Wizard portrays AI as powerful, yet, unpredictable and inexplicable; and Pinocchio imagines AI as gaining free will, learning from mistakes, and possibly harming humans. These images of AI provide a novel framework for understanding how AI practitioners perceive and shape AI solutions. Our findings and recommendations will assist AI practitioners, companies, and users with a shared vocabulary and understanding to explicitly and critically examine the intended and unintended impacts of AI solutions on human society, contributing to more responsible and human controlled AI design and use. • Technological advancement, commercial investment, and public interest in AI continues to accelerate. • Yet, there is little insight into how everyday AI practitioners perceive AI and its impact on society. • We inductively identified six Images of AI, ways in which AI practitioners view AI. • AI companies and practitioners can use our findings to develop more responsible AI solutions. • AI users can use our vocabulary and framework to better control their AI usage.