arXiv:2602.21889v2 Announce Type: replace-cross Abstract: Predictions from ML models support human decision making in several fields, including high-stakes ones such as healthcare and the judiciary. Yet, we still lack a clear understanding of how decision makers learn from ML-based decision support (ML-DS). In this paper, we introduce a general computational framework, the 2-Step Agent, to capture this process. As a prediction from an ML model contains information about the training data, a prediction can also be used for inference. Our framework models (i) how a prediction for a new observati
The proliferation of ML models in high-stakes decision-making fields necessitates a deeper understanding of human-AI interaction for effective and safe integration, especially as AI advances rapidly.
This framework provides a computational model for how humans learn from and integrate AI decision support, which is critical for designing more effective, trustworthy, and ethically aligned AI systems in high-impact domains.
Our understanding of the dynamic between human decision-makers and AI tools is refined, moving from simple prediction use to a more nuanced view of AI-as-inference, informing the design of future 'AI agents'.
- · AI developers
- · Healthcare providers
- · Judiciary systems
- · Human-AI collaboration platforms
- · Opaque AI systems
- · Non-adaptive decision support tools
- · Human-AI interaction models lacking learning mechanisms
Improved design and adoption of AI decision support systems across critical sectors.
Increased trust and better outcomes in high-stakes areas due to more effective human-AI synergy.
The acceleration of autonomous AI agents capable of truly adaptive interaction with human oversight.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG