Collaborative Intelligence: A scoping review of current applications

Humans and artificial intelligence (AI) systems have complementary strengths. This complementarity creates the potential to improve performance by combining inputs from human and AI on a common task or goal. A systematic review of academic and grey literature was conducted to investigate whether real-world examples of ‘collaborative intelligence’ could be identified. Applications utilising collaborative intelligence had to have (1) complementarity (i.e., the collaboration improves performance beyond that which could be achieved by the human or the AI alone), (2) a shared objective and outcome, and (3) sustained, task-related interaction between human and AI. The literature search yielded 1250 documents but only 16 applications meeting these criteria were identified. Most collaborative applications were at the prototyping stage but they could perform a variety of types of work (creative, industrial, healthcare, emergency response and knowledge work) and delivered a range of benefits (efficiency, creativity and safety). In most applications the AI had a virtual presence but there were examples where the AI also had a cyber-physical form. These early applications reveal the potential for collaborative intelligence to be applied in a range of domains and formats.