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

Cedric Faas, Richard Bergs+3 more|arXiv (Cornell University)|Oct 10, 2024

Impact of AI and Big Data on Business and Society4 citations

TL;DR

This study investigates how restricting choice in AI-supported decision-making affects perceived autonomy, motivation, and performance. In a simulated drone oversight task with 274 participants, limiting AI-recommended actions to one significantly improved decision accuracy but reduced perceived autonomy and task meaningfulness—effects that worsened over time, while higher autonomy predicted better performance in multi-choice conditions.

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.

Motivation & Objective

• To examine how restricting choices through AI decision support affects perceived autonomy and motivational variables in human-AI collaboration.
• To investigate the long-term consequences of reduced choice on psychological needs and task performance in AI-supported decision-making.
• To explore the relationship between perceived autonomy, task meaningfulness, and decision accuracy in human-AI oversight roles.
• To highlight the importance of integrating motivational and psychological needs into human-AI interaction design beyond cognitive optimization.
• To provide empirical evidence on the trade-offs between performance gains and motivational costs when limiting user choice in AI systems.

Proposed method

• Conducted a between-subject experimental study with 274 participants in a simulated drone oversight task.
• Participants faced 10 critical decision scenarios where an AI recommended 6, 4, 2, or 1 selectable actions.
• Used a controlled experimental design to manipulate the number of selectable actions (1, 2, 4, or 6) while keeping AI recommendations accurate.
• Measured perceived autonomy, task meaningfulness, motivation, and decision accuracy across time and conditions.
• Applied work design theory (e.g., Hackman & Oldham, 1975; Ryan & Deci, 2000) to frame psychological needs and motivational variables.
• Collected self-report data on perceived autonomy, meaningfulness, and motivation, with objective performance measured by decision accuracy.

(a) One Selectable Actions

Experimental results

Research questions

• RQ1 How does restricting the number of selectable actions through AI support affect participants’ perceived autonomy in an oversight task?
• RQ2 What is the impact of choice restriction on perceived task meaningfulness and intrinsic motivation in human-AI decision-making?
• RQ3 How does perceived autonomy relate to decision accuracy across different levels of choice restriction?
• RQ4 Do the effects of choice restriction on autonomy and motivation change over time in repeated decision scenarios?
• RQ5 To what extent do motivational variables mediate the relationship between choice restriction and decision performance?

Key findings

• Restricting AI-supported choices to one action significantly reduced participants’ perceived autonomy compared to conditions with two or more selectable actions.
• The reduction in perceived autonomy was accompanied by a significant decline in perceived task meaningfulness, particularly in the one-choice condition.
• The negative effects on autonomy and meaningfulness intensified over time, indicating long-term motivational costs of restricted choice.
• In conditions with multiple selectable actions, participants who reported higher perceived autonomy demonstrated significantly better decision accuracy.
• Despite improved accuracy in the one-choice condition (due to a perfect AI), the trade-off in psychological well-being and motivation undermines long-term collaboration effectiveness.
• The findings suggest a reciprocal relationship between motivation and performance, where high motivation enhances performance and high performance may reinforce motivation, but choice restriction disrupts this cycle.

(b) Two Selectable Actions