Reimagining workflows to take advantage of the complementary capabilities of humans and AI and create effective, efficient and enjoyable collaborative intelligence systems.
AI does not operate in isolation; it is used within a system of interconnected tasks (a workflow) by skilled and knowledgeable users. To make sure humans and AI can collaborative effectively, we need to consider how the AI will be integrated into the broader workflow: what tasks will the AI complete and what tasks are best performed by humans? How will information flow between humans and AI, and how can we ensure their work contributes to the same goal? Poorly integrated AI may result in slow or inaccurate workflows, and users may feel confused, bored or frustrated. Realising the full benefits of CINTEL may involve creating entirely new workflows, to ensure optimal performance of the human-AI team.
This project draws on psychology and human factors, using a combination of applied and experimental research. Working with CINTEL collaborators, we document existing workflows in real scientific use cases, to identify the expertise scientists bring to their work and the challenges that AI can help with. This knowledge helps shape the design of new workflows.
We employ experimental psychology methods to better understand the principles of effective human-AI teamwork. This research evaluates different ways of constructing workflows and allocating tasks to humans or AI, with a focus on optimising performance of the human-AI team.
Insights from this project inform the development of best practice guidelines for human-AI workflows, and a methodology that can be applied by developers and scientists to create new workflows within their own domain.