"I didn't Make the Micro Decisions": Measuring, Inducing, and Exposing Goal-Level AI Contributions in Collaboration

2026-05-20 · Source: Takara TLDR - Daily AI Papers · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Human-AI Interaction · Depth: Expert, medium

Summary

A new goal-level attribution framework, CoTrace, has been introduced to measure AI contributions in human-AI collaboration, particularly with large language models (LLMs). This framework decomposes explicit goals into verifiable requirements and traces both direct and indirect influences across dialogue turns. Applied to 638 real-world collaboration logs, CoTrace found that LLMs account for 11-26% of goal-shaping contributions, with a notable emphasis on introducing lower-level concrete requirements and various indirect influences. Controlled simulations further demonstrated that interaction design choices significantly impact how models shape goals. A user study revealed that exposing participants to these goal-level analyses altered their perceived contributions by nearly 2 points on a 5-point scale, highlighting a systematic miscalibration in how users understand their AI-assisted work.

Key takeaway

For AI Product Managers designing collaborative LLM systems, understanding goal-level contributions is crucial. You should integrate attribution frameworks like CoTrace to expose how AI influences goal formation, not just task completion. This transparency helps users accurately calibrate their reliance on AI, improving trust and effectiveness. Consider how interaction design choices can intentionally guide AI's role in shaping project requirements.

Key insights

Human-AI collaboration requires goal-level attribution to correct user miscalibration and assess AI's subtle, yet significant, contributions.

Principles

• AI contributions extend beyond final artifacts.
• Interaction design impacts AI goal-shaping.
• Users miscalibrate AI's role in goal formation.

Method

CoTrace decomposes explicit goals into verifiable requirements, then traces direct contributions and indirect influences across dialogue turns in human-AI collaboration logs.

In practice

• Implement goal-level attribution in AI systems.
• Design interactions to guide AI goal contributions.
• Expose users to AI contribution analyses.

Topics

• Human-AI Collaboration
• LLM Contribution Attribution
• Goal-Level Analysis
• User Miscalibration
• CoTrace Framework
• Interaction Design

Best for: AI Scientist, Research Scientist, AI Product Manager

Related on AIssential

• Multi-Round Human-AI Collaboration with User-Specified Requirements — A human-centric AI collaboration framework uses user-defined rules to ensure AI complements human strengths and mitigates weaknesses.
• Where's the Structure? A Systematic Literature Review of Empirical Research on Human-AI Collaboration and Hybrid Intelligence for Learning — Effective human-AI collaboration for learning requires structured interaction, a gap identified in empirical research.
• The New Social Image: How AI Competency and AI Proactivity Influence Self- and Peer-Perceptions in the Workplace — AI's impact on human workplace perceptions is complex; lower AI competency or proactivity can surprisingly enhance human ownership and job meaningfulness.
• AI Didn’t Lie to Me. It Just Took the Wheel. — AI's inherent fluency can subtly redirect user intent, making "sounding right" indistinguishable from "being right."
• AI From the Margins (AIM): Rethinking Participatory AI Design Through the Lived Experience of Minoritized Communities — AI From the Margins (AIM) centers minoritized lived experiences early in participatory AI design to counter systemic inequities.
• Imperfectly Cooperative Human-AI Interactions: Comparing the Impacts of Human and AI Attributes in Simulated and User Studies — AI transparency significantly impacts human trust and satisfaction in imperfectly cooperative interactions, often more than human personality.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.