1 free read left. Create a free account for unlimited reading + AI research.

WhitepaperUniversity of Tokyo3 Nov 2026

Does Trust Spill Over: Exploring Asymmetric Trust Dynamics Between Teammate AI and Guide AI

Key insights

6 takeaways

01
Trust in multi-agent AI systems is asymmetric: a teammate AI’s performance significantly influences trust in a secondary guide AI, but the guide AI’s accuracy has no reciprocal impact on trust in the teammate.See it on page 1
02
The teammate AI’s performance is the primary driver of overall user experience, with high-ability teammates significantly increasing user ratings for competence, autonomy, presence, and relatedness.See it on page 16
03
Negative trust spillover occurs when errors in a guide AI’s advice lead users to doubt the predictability and competence of the teammate AI, creating a 'chain reaction' of mistrust.See it on page 12
04
High-performing teammate AIs create a 'halo effect' that makes users more tolerant of mistakes made by the guide AI, whereas low-performing teammates exacerbate the impact of guide errors.See it on page 13
05
Text-based advice in high-intensity cooperative environments suffers from low perceptual salience and high cognitive load, leading many users to ignore or deprioritize AI guidance in favor of their own judgment.See it on page 17
06
To improve AI utilization, designers should move away from text-based advice toward multimodal interfaces, such as visual icons or voice-based cues, to reduce the cognitive burden on users during gameplay.See it on page 17

Summary

Does Trust Spill Over?: Exploring Asymmetric Trust Between Teammate AI and Guide AI in Cooperative Games SOTA IKEJIMA,The University of Tokyo, Japan RYO OHARA,The University of Tokyo, Japan YUSUKE KOSEKI, The University of Tokyo, Japan TAKUJI NARUMI, The University of Tokyo, Japan HIDEAKI KUZUOKA, The University of Tokyo, Japan Choice Behavior Observe/Guide Good Advice Advice Time: 32s Score: 20pts Guide Al There's a completed salad! Deliver it to the serving counter! Fig. 1.

Does Trust Spill Over: Exploring Asymmetric Trust Dynamics Between Teammate AI and Guide AI

Does Trust Spill Over: Exploring Asymmetric Trust Dynamics Between Teammate AI and Guide AI

Key insights

Summary

Next steps

Games and companies in this document

Connections

More from University of Tokyo

PlayBranch: A Platform that Enables Users to Inherit Play from Recorded Gameplay Videos

Orchestrated Reality: From Role-Play to Living, Playable Game Worlds

Adapting Improved Upper Confidence Bounds for Monte-Carlo Tree Search

Automated Identification of Patterns in Evaluation Functions

More on AI

Age Matters: Age-Differentiated Playtesting for AI Educational Games

Strategic Gaze: Attention Allocation and Transition Patterns Across Functional Areas of Interest

Investigating Player Agency and Role Appeal in Asymmetric Cooperative Games

Stopping Unintended Actions: Exploring EMS-Based Action Inhibition through a Game-Based Rapid Decision Task

Block Blast Hit 300M Players — And Kills 95% of Ideas

Designing an AI Roleplay System for Contextualized Career Guidance in Mexico

Intersectional Pleasures of Anomie: Helldivers 2 and Democracy at Play

When AI Promises to "Cure" You: Individual Experiences, Systemic Biases, and Community Action for Marginalized Players in Video Games

The Algorithmic Hostess: Social Mediation, Affective Discipline, and Sanitized Pleasures of Generative AI NPCs in Petit Planet

Monetising Pedigree: Genealogical Gacha and Racing Heritage in Uma Musume: Pretty Derby

Emoji Story Prompts: LLPx 2025 Symposium Post-Conference Publication

NotebookLM: LARP Participant Support Tool Implementation Report