Modeling Human Behavior Part II -- Cognitive approaches and Uncertainty

Cognitive limitations shape decision-making: The paper emphasizes that human reasoning is constrained by limited cognitive resources, and AI systems designed to model human behavior must account for these limitations rather than assuming humans are perfect rational actors. This is important for AI systems that interact with or make decisions about people, as they need to reflect realistic human capabilities.

Cognitive biases are predictable and modelable: The research demonstrates that human biases and heuristics—mental shortcuts people use when reasoning—can be systematically represented and incorporated into AI models. Understanding these patterns helps create AI systems that better predict human choices and can identify when humans might make predictable errors.

Uncertainty representation matters for realistic modeling: The paper highlights that effective human behavior models must explicitly represent and utilize uncertainty, rather than treating human decision-making as deterministic. This is crucial for responsible AI because it acknowledges that both humans and AI systems operate with incomplete information, which should be transparently reflected in how AI systems make predictions about human outcomes.

Cognitive architectures provide structured frameworks: By studying formal cognitive architectures and how humans actually think, researchers can build AI systems with more transparent and interpretable decision-making processes. This supports responsible AI development by making it clearer how and why AI systems reach conclusions about human behavior.