AI Optimizes Outputs While Destroying the Will to Use Them

AI tools can generate objectively better goals than users write themselves, but motivation depends on psychological ownership, not just goal quality. Does optimization come at the cost of commitment?

Method: LLM-generated goals scored higher on SMART criteria (d = 2.26) but participants reported lower psychological ownership (d = 1.38), commitment (d = 1.19), and perceived importance (d = 1.13). At two-week follow-up, 72.8% of self-authored participants acted on two or more goals versus 46.6% in the LLM condition. Mediation analysis confirmed ownership—not objective quality—drove every downstream motivational and behavioral outcome.

Caveats: Two-week follow-up only. Longer-term adaptation or ownership recovery mechanisms unexplored.

Reflections: Can hybrid authorship models (AI suggests, user edits) preserve ownership while improving goal quality? · Does repeated exposure to AI-authored goals eventually restore ownership through habituation? · Do ownership effects generalize to other identity-relevant tasks like value articulation or career planning?

Agentic AI handles routine customer service autonomously, but failures create both cognitive and emotional consequences. How effective are human interventions when AI escalates?

Method: AI deployment reduced chat duration and had limited effects on retrial rates but substantially lowered ratings for AI-eligible chats. Human intervention preserved service quality in algorithm-triggered technical escalations but failed in emotional escalations. Workers sent fewer messages, contributed a smaller share of chat rounds, and showed less proactivity in emotional escalations. Early intervention was essential for sustaining high post-escalation effort.

Caveats: Single platform (Alibaba Taobao). Worker training, incentives, and cultural context may not generalize to other customer service operations.

Reflections: Can explicit emotional escalation training or incentive redesign restore worker engagement in affective failures? · Do customers perceive emotional escalations as AI failures even when human intervention is immediate? · What handoff UI patterns signal emotional urgency effectively enough to overcome worker disengagement?

Text-to-3D generative AI creates models easily but provides limited control over spatial structure, making precise geometric composition impossible. How do you give users spatial authority while keeping generative convenience?

Method: MiXR lets users extract segments from captured real-world objects and assemble them through direct 3D manipulation in XR, then delegates geometric refinement to generative models. In a controlled study (N=12), participants rated MiXR designs as significantly closer to target, felt more in control, and experienced lower cognitive workload versus generative composition baseline.

Caveats: Small sample (N=12). Generalizability to non-XR contexts or professional design workflows unverified.

Reflections: Does the approach scale to complex assemblies with dozens of harvested segments? · Can the hybrid workflow transfer to 2D design tools where spatial manipulation is less direct? · How does harvesting from real objects versus digital libraries affect creative outcomes?