The model of agent systems is based on the constraint nets of Zhang and Mackworth [1995] and Rosenschein and Kaelbling [1995]. Walter [1950] presents a simple tortoise agent that has interesting behavior, and Walter [1951] describes a version that learns. Luenberger [1979] is a readable introduction to the classical theory of agents interacting with environments. Turtle Geometry, by Abelson and DiSessa [1981], investigates mathematics from the viewpoint of modeling simple reactive agents.

The hierarchical control is based on Albus [1981] and the subsumption architecture of Brooks [1986]. Simon [1996] argues for the importance of hierarchical control.

Kahneman [2011] provides compelling evidence for distinguishing two modes of human thought: fast, instinctive and, emotional versus slow, deliberate, and rational, which he calls Systems 1 and 2 to avoid oversimplification.

For more detail on agent control, see Dean and Wellman [1991], Latombe [1991], and Agre [1995]. The methodology for building intelligent agents is discussed by Haugeland [1985], Brooks [1991], Kirsh [1991b], and Mackworth [1993].

Qualitative reasoning is described by Forbus and Hinrich [2017] and Forbus [2019].

The moral machines experiment is described by Awad et al. [2018, 2020] and Bonnefon [2021].