Simulation of Human Memory

欣茨曼之前由美国国家科学基金会支持的研究表明，一个简单的人类记忆理论，基于一些核心假设，可以解释在人类学习和记忆实验中发现的惊人数量的经验现象。该理论的基本假设已经在计算机程序中形式化，这些程序已经在学习和记忆任务中进行了测试，类似于人类参与实验的任务。这样的模拟工作对认知理论的发展是有价值的帮助，特别是涉及并行处理和可变性等因素的理论，其结果很难直观地预测。例如，欣茨曼模拟模型的基本假设只涉及两个基本变量的运行过程，即重复和相似。然而，该模型解释了几个以前被认为是独一无二的现象，需要他们自己特别的解释。这项研究将包括三类活动：(A)模拟模型将以两种方式扩展。首先，到目前为止，本模型主要应用于对不同事件的记忆和抽象类别的学习，将应用于联想学习和回忆领域。第二，将开发不同版本的模式，以处理现有模式不足以处理的问题。这些包括从记忆中提取单词的意思，以及先前熟悉对学习的促进作用。(B)几个实验将测试该模型作出的关于抽象类别的学习、重复和相似性对先前遇到的项目的识别的影响以及记忆提取潜伏期的新预测。(C)将目前的记忆模型与文献中的其他模型进行比较。人们的兴趣将主要集中在确定是否存在其行为本质上与当前模型相同的“连接主义”类型的简单模型。这一点很重要，因为一些人声称连接主义模型的一个优势是，它们在解剖学或生理学上是可信的。这一理论对学习、记忆和认知过程有着广泛的影响。因此，它可以在教育和培训、改善记忆力、治疗记忆缺陷以及基于记忆的判断的可靠性等领域产生实际的见解。

Previous NSF-supported research by Hintzman has shown that a simple theory of human memory, based on a few core assumptions, can account for a surprising number of empirical phenomena uncovered in experiments on human learning and memory. The basic assumptions of the theory have been formalized in computer programs, which have been tested on learning and memory tasks analogous to those given to humans participanting in experiments. Such simulation work is a valuable aid in the development of theories of cognition, particularly theories involving factors such as parallel processing and variability, whose consequences are difficult to anticipate intuitively. The basic assumptions of the Hintzman simulation model, for example, concern processes underlying the operation of just two fundamental variables, repetition and similarity. However, the model accounts for several phenomena that previously had been thought to be unique, and to require their own ad hoc explanations. The research will consist of three types of activity: (a) The simulation model will be expanded in two ways. First, the present model, which has so far been applied primarily to memory for distinct events and the learning of abstract categories, will be applied to the domains of associative learning and recall. Second, different versions of the model will be developed to handle problems for which the present model is inadequate. These include retrieval of the meaning of a word from memory, and the facilitative effect of prior familiarity on learning. (b) Several experiments will test new predictions made by the model concerning the learning of abstract categories, the effects of repetition and similarity on recognition of previously encountered items, and memory retrieval latencies. (c) Comparisons will be made between the present memory model and other models in the literature. Interest will focus primarily on determining whether a simple model of the "connectionist" type exists whose behavior is essentially the same as that of the current model. This is important because an advantage claimed for connectionist models by some is that they are anatomically or physiologically plausible. The theory has broad implications for learning, memory, and cognitive processes. As such, it could yield practical insights in the fields of education and training, memory improvement, treatment of memory defects, and the reliability of judgments based on memory.