Towards the design of user-centric strategy recommendation systems for collaborative Human-AI tasks

Area of Science:

• Human-Computer Interaction
• Artificial Intelligence
• Cognitive Science

Background:

• Artificial intelligence (AI) is increasingly used for complex collaborative tasks.
• Personalized recommendation systems are common in e-commerce and social media.
• Understanding user preferences is key for effective AI-human teamwork.

Purpose of the Study:

• Investigate factors for user-centric AI strategy recommendation systems.
• Measure user preferences across different strategy recommendation modalities.
• Analyze the impact of personality on system choice.

Main Methods:

• Human-subjects experiment with 60 participants.
• Evaluated four recommendation types: single, similar, diverse, and all strategies.
• Measured user preferences and system perceptions.

Main Results:

• Conscientiousness significantly impacts preference for specific recommendation systems (p < 0.01).
• Higher perceived alignment with user preferences increases perceived intelligence (p < 0.01).
• Greater alignment also leads to higher perceived usability (p < 0.01).

Conclusions:

• User personality traits are critical in designing AI strategy recommendation systems.
• System alignment with user preferences enhances perceived intelligence and usability.
• Findings inform the development of more effective user-centric AI collaboration tools.