Beyond Cognitive Appraisal: An AI Delegation Framework for Human-AI Collaborative Intelligence

Description

Artificial Intelligence (AI) is disrupting and reinventing business processes. Its role has shifted from traditional decision support to a more nuanced reality of decision automation. While AI has shown its ability to complement or even replace humans in complex analytical decisions, human-AI collaboration, rather than either one dominant in the marketplace alone, has been predicted as the future trend. To date, little is known about how humans distribute rights and responsibilities to AI by delegating operations management decisions, which has been regarded as one of the main barriers to human-AI collaborative intelligence.This project draws upon the attribution theory and cognitive appraisals to explore the delegation relationship between a human manager (as a delegator) and an AI system (as a proxy). The attribution theory-based cognitive appraisal, which is conducted actively and continuously by the delegator through observing the proxy’s behavior and job outcomes, can shape the delegator’s perception of the proxy’s endowment. When a delegation decision should be made, the previously perceived endowment of the proxy will be recalled. Cognitive appraisal has been adopted to theorize effective delegation between humans. However, few studies have investigated how human managers conduct appraisals of autonomous AI to make delegation decisions in the context of human-AI collaboration.We intend to extend the cognitive appraisals in an AI delegation framework for a better understanding of human managers’ willingness to delegate their routine operational tasks to an AI system. We propose that human managers’ perceived AI endowment is developed from both subjective appraisals (i.e., whether AI can mimic human decisions and meet managers’ subjective expectations) and objective appraisals (i.e., whether AI can accomplish the task and achieve a better sales performance). In addition, human managers’ subsequent appraisal process is biased toward the existing category assigned to AI. The delegation outcomes can further shape managers’ confidence in their judgment of AI from a repeated appraisal in the delegation feedback loops.To this end, we have collected real-world field data of daily delegation decisions and delegation outcomes from 1,006 store managers during the study period from January 2017 to September 2019, including AI’s daily operations decisions, managers’ delegation decisions, performance reports, and managers’ assessment reports. The AI system was implemented in March 2018. We intend to use this dataset to examine the underlying mechanism between managers’ cognitive appraisals of AI, task attributes, and delegation willingness. Considering the trend of human-AI collaboration in decision automation will continue, the deliverable of this project will bring significant implications to companies that use or plan to use AI systems for decision automation by directing the fastest pathway to human-AI collaborative intelligence.