Metacognitive Agents for Ethical Decision Support: Conceptual Model and Research Roadmap
It addresses ethical decision-making challenges for AI systems, but is conceptual and incremental as it outlines a research roadmap without implementation.
The paper tackles the ethical value-action gap, where intentions and actions diverge due to obstacles and biases, by proposing a roadmap to develop metacognitive assistant agents for value-aligned decision support.
An ethical value-action gap exists when there is a discrepancy between intentions and actions. This discrepancy may be caused by social and structural obstacles as well as cognitive biases. Computational models of cognition and affect can provide insights into the value-action gap and how it can be reduced. In particular, metacognition ("thinking about thinking") plays an important role in many of these models as a mechanism for self-regulation and reasoning about mental attitudes. This paper outlines a roadmap for translating cognitive-affective models into assistant agents to help make value-aligned decisions.