Give Me a Choice: The Consequences of Restricting Choices Through AI-Support for Perceived Autonomy, Motivational Variables, and Decision Performance

TL;DR

It is proposed that effective human-AI collaboration requires broader consideration of human needs (e.g., autonomy) that affect motivational variables (e.g., meaningfulness) that affect motivational variables (e.g., meaningfulness) that affect motivational variables (e.g., autonomy).

Abstract

Design optimizations in human-AI collaboration often focus on cognitive aspects like attention and task load. Drawing on work design literature, we propose that effective human-AI collaboration requires broader consideration of human needs (e.g., autonomy) that affect motivational variables (e.g., meaningfulness). In a simulated drone oversight experiment, participants (N=274, between-subject) faced 10 critical decision-making scenarios with varying levels of choice restrictions with an AI recommending only 1, 2, 4 or all 6 possible actions. Restricting participants to one selectable action improved task performance (with a perfect AI) but significantly reduced perceived autonomy and work meaningfulness, and these effects intensified over time. In conditions with multiple action choices, participants with higher perceived autonomy performed better. The findings underscore the importance of considering motivational factors to design successful long-term human-AI collaboration at work.

Figures (5)

• Figure 1: Differences in the decision-making user interface depending on experimental conditions.
• Figure 2: Screenshots of the Drone Monitoring Interface. On the top left, a screenshot of the demo task is shown, where participants see a video recorded by the drone and information about the current status of the drone. In the top middle, the legend of all icons which were used in the interface is shown. This legend was shown to the participants during the entire video of the drone. Ten seconds before the participants had to make a decision, the drone entered a critical situation, which was indicated by an auditory signal and highlighted icons for critical values, as shown in the top right. After these ten seconds, the participants saw six possible actions, and depending on the experimental condition some of them were grayed out indicating they were not available as seen in the bottom screenshot of the TwoSelectable Actions condition. In this demo task, the critical situation was the drone approaching trees which could be resolved by going up.
• Figure 3: Differences in perceived autonomy, meaningfulness, task motivation, and task load on a 7-point Likert Scale between the experimental conditions
• Figure 4: Differences in Decision Accuracy in Percentage and Decision Time in milliseconds between the Experimental Conditions
• Figure 5: Differences in perceived autonomy, meaningfulness, task motivation, and task load over time in each condition on a 7-point Likert Scale