The Computational Problem of Prospective Memory Retrieval

Although the need to remember to do things in the future is important and ubiquitous in human life, this phenomenon of prospective memory has received little attention in the cognitive modeling community. One roadblock is the lack of a computational definition of the phenomenon. We propose that prospective memory is fundamentally a problem of circular knowledge dependencies. This framework allows us to explain the success of human preemptive, spontaneous retrieval, and noticing-plus-search strategies. It also provides a clear mapping of the problem onto the capabilities of cognitive architectures. We interpret previous ACT-R and Soar models of prospective memory in this light, and identify areas of architectural development that would lead to more complete models of prospective memory.