Research map / Large language models / AI and society

Human-in-the-loop or AI-in-the-loop? Automate or Collaborate?

AAAI 2025 · Large language models › AI and society · approach: Human-centered trust and agency · 47 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

A blue-sky position paper arguing that many so-called human-in-the-loop systems are really AI-in-the-loop systems where the human holds control, and that evaluation should be human-centric accordingly.

Problem

Human-in-the-loop systems are often evaluated purely on AI/ML performance metrics, ignoring that in many deployments the human is the decision-maker and the AI only supports them.

Approach

Distinguishes HIL (AI in control, humans supply labels or guidance) from AI-in-the-loop (AI2L: human in control, AI assists), arguing they differ in control, source of bias, and evaluation, and should be designed and evaluated separately.

Surprising findings

• Current foundation models lend themselves more to HIL settings where AI is the primary actor, rather than to collaborative AI2L settings.

Limitations

• No empirical validation of the proposed AI2L perspective; argument is conceptual.
• Examples in the domain table are illustrative rather than measured.

Experimental setup

Builds on: active learning interactive machine learning machine teaching

Closest papers by meaning

• Formalising Human-in-the-Loop: Computational Reductions, Failure Modes, and Legal-Moral Responsibility (ICLR 2026 · similarity 0.85)
• Human Expertise in Algorithmic Prediction (NeurIPS 2024 · similarity 0.83)
• Balancing Humans and Machines: A Study on Integration Scale and Its Impact on Collaborative Performance (AAAI 2024 · similarity 0.83)
• Position: Towards Bidirectional Human-AI Alignment (NeurIPS 2025 · similarity 0.82)
• Designs for Enabling Collaboration in Human-Machine Teaming via Interactive and Explainable Systems (NeurIPS 2024 · similarity 0.82)
• When Models Know More Than They Can Explain: Quantifying Knowledge Transfer in Human-AI Collaboration (NeurIPS 2025 · similarity 0.82)