ABSTRACTConventional cartography practices situate a cartographer in a computer environment in which choices are made about a limited number of map characteristics. A map is then produced using these choices. Such maps are often judged inadequate and an alternative will be generated, typically by changing a parameter value. This does not always lead to a satisfactory outcome, particularly when, as is increasingly common, novices are involved in mapmaking. The purpose of this paper is to initiate a paradigm change in statistical cartography, one in which the production of maps switches from a sequence of actions taken by a mapmaker to a process of specifying criteria that are used to create maps using intelligent agents. A solution is then selected that satisfices the elicited criteria and others derived from cartographic theory and praxis. The cartographer’s role in this paradigm shifts from involvement in a sequence of low-level software-mediated tasks to a higher level. This active symbolism approach is described and then illustrated using the production of dot maps to illustrate how collections of intelligent design agents are used to generate emergent alternatives that can be evaluated against design criteria and how deep learning can be introduced as part of the design process.