What it takes to control AI by design
Human learning
Bibliographic Data
| ID | 20397980 |
|---|---|
| Authors | Dov Te’eni (0000-0002-5629-8614, Tel Aviv University, corresponding author), Inbal Yahav (0000-0002-1513-017X, Tel Aviv University), David G Schwartz (0000-0002-2125-2069, Bar-Ilan University), David Schwartz (0000-0002-3989-8546) |
| Year | 2026 |
| Volume | 41 |
| Issue | 1 |
| Pages | 237-250 |
| Publication date | 2026-01-01 |
| Open Access | Yes |
| Type | ARTICLE |
| Venue | AI & Society (JOURNAL) |
| Journal identifiers | ISSN: 0951-5666 • E-ISSN: 1435-5655 |
| Publisher | Springer Science and Business Media LLC (PUBLISHER) |
| DOI | 10.1007/s00146-025-02401-y |
| OpenAlex | W4413472956 |
| Language | EN |
| References cited | 41 |
Experts in government, academia, and practice are increasingly concerned about the need for human oversight in critical human–AI systems. At the same time, traditional control designs are proving inadequate to handle the complexities of new AI technologies. Incorporating insights from systems theory, we propose a robust framework that elucidates control at multiple levels and in multiple modes of operation, ensuring meaningful human control over the human–AI system. Our framework is built on continual human learning to match advances in machine learning. The human–AI system operates in two modes: stable and adaptive, which, in combination, enable the effective use of big data and the learning necessary for effective control and adaptation. Each system level and mode of operation requires a specific control-feedback loop, and all controls must be aligned for performance and values with the higher system level to provide human control over AI. Applying these ideas to a human–AI decision system for text classification in critical applications, we demonstrate how a method we call reciprocal human–machine learning can be designed to facilitate an adaptive mode and how oversight can be implemented in a stable mode. These designs yield high and consistent classification performance that is unbiased and closely aligned with human values. It ensures effective human learning, enabling humans to stay in the loop and stay in control. Our framework provides spadework for a model of control in critical AI decision systems operating in volatile environments, where humans continue to learn alongside the machine
Art · Human–computer interaction · Knowledge management · Performing arts · Visual arts · Computer Science · Engineering · Ethics and Social Impacts of AI · Explainable Artificial Intelligence (XAI · Human-Automation Interaction and Safety · Artificial Intelligence
Human Error
Human-Centered Artificial Intelligence
Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead
Meaningful Human Control over Autonomous Systems
Regularization Paths for Generalized Linear Models via Coordinate Descent
Organizing and the Process of Sensemaking
Corporate social responsibility and green supply chain efficiency
Explainable AI in the military domain
A hybrid mixed methods design of qualitative enhancement and reciprocal feedback loop for augmented text classification
What machines shouldn’t do
Friendly AI will still be our master. Or, why we should not want to be the pets of super-intelligent computers
Artificial Intelligence and Black‐Box Medical Decisions
Prevalent Sentiments of the Concept of Jihad in the Public Commentsphere
The design of inquiring systems
Human-Machine Reconfigurations
An introduction to cybernetics
| Highly cited | No |
|---|