AI-assistance to decision-makers: evaluating usability, induced cognitive load, and trust’s impact

We designed a randomized-controlled study with 80 participants to investigate the effects of an AI assistant on the activity and processes implemented by users in their decision-making tasks. For that purpose, we will examine several aspects of decision-making in problem-solving situations in five experimental conditions resulting from the combination of the following factors: AI assistance (with vs. without), information related to the reliability of the assistant’s proposals (yes vs no) and cognitive load induced variation through a dual task (with vs without). We plan to collect profile data, heart-rate variables, task efficiency and perceived usability, cognitive load, and trust. We are currently finalizing the prototype to conduct pretests.