Designing human compatible AI
Secondmind has developed the world’s first principled AI platform for generalised decision-making. It learns autonomously and makes decisions in order to achieve goals in high-dimensional, dynamic systems like financial markets, smart cities, online games and logistics. Secondmind invited Dovetailed to conduct comprehensive user research, and we have worked closely with their team to understand how humans might interact with future AI-based systems to achieve desirable outcomes.
Qualitative and quantitative data collection
Of particular interest has been how to build trust in AI so that humans are comfortable acting on computer-driven instructions and how to motivate people to act in a way which may seem counter-intuitive. Our research and analysis of unmet needs has shaped a vision for AI
user-experience design, still relatively unexplored in human-computer interaction (HCI).
We conducted 1-to-1 interviews with 15 participants who were recruited from five different industries, identified by Secondmind as being of high strategic importance for the company. User questionnaires were developed in a highly collaborative manner with the internal development team to fully understand the operational demands of particular industries and uncover potential insights into how AI-based systems could add value. Discussion of existing workflows, daily routines and responsibilities helped to shape user personas and visionary user-journeys which in turn informed later UX design work and wireframes.
The UX of AI is a continual work-in-progress as it is a vast new space to explore. Our research, design and user studies have helped Secondmind to answer fundamental questions which will enable their vision. This vision-setting informs their companies on-going mission to deliver the best of AI through intelligently designed solutions.
It’s been great to understand and identify what real world pain points our AI algorithms will be helping to solve in the near future
— Head of Group