CompCog: The Role of Control Processes in the Dynamics of Episodic Memory Search

记忆将我们的经历联系在一起。记住一件事会让我们想起另一件事。这可能是一种祝福，就像一首歌将你带回过去，重温一段美好的记忆。或者这可能是一种诅咒，就像当你看到一位同事时，会让你想起错过了最后期限。控制脑海中浮现的记忆会影响一个人的幸福感。这项研究将研究为什么某些记忆会联系在一起，以及允许我们搜索特定记忆而避开其他记忆的过程。现有的研究经常使用任务，通过让参与者只学习随机单词来简化现实世界的记忆。我们将研究将现实世界的复杂性添加到任务中如何影响行为。我们工作的另一个特点是使用计算机模型来模拟假想的记忆搜索背后的认知机制。这些模拟让我们可以评估记忆理论。这项工作还将有助于培养下一代科学家。我们将在我们的实验室培养研究生，并为当地高中开发一个研讨会。学生将收集和分析数据。这将向他们介绍记忆科学和计算机编程。研究有两个科学目标：目标1是阐明基于时间的联想和基于意义的联想之间相互作用的计算基础。发生在时间附近的事件往往通过意义联系在一起，但典型的自由回忆任务缺乏基于意义的物品之间的关系。因此，受试者在很大程度上依赖基于时间的联想来指导记忆搜索：回忆一个单词，并以此为线索回忆起附近研究过的另一个单词。但最近的数据显示，当列表包含强烈的基于意义的关联时，受试者几乎放弃了基于时间的关联。为了确定时间和语义信息如何相互作用来指导记忆搜索，我们的实验将改变列表中语义关联的数量和性质。我们的模拟将测试这样一个假设，即这种交互可以通过在网络的不同层中表示语义相似性和时间邻近性来解释。目标2是阐明使控制过程能够改变记忆编码和搜索的动力学的计算原理。我们经常在没有尝试的情况下形成新的记忆，后来发现线索会自发地把它们带到脑海中。其他时候，我们故意对信息进行编码，以实现某些目标。但大多数实验室记忆任务并不完全像这些情况：受试者故意记住项目，但只是为了死记硬背，而不是为了一个特定的目标。最近的数据表明，偶然编码和目标定向编码都显著改变了受试者对基于时间的联想的依赖。这表明，允许记忆系统适应不同任务需求的控制过程，是记忆如何组织的关键决定因素。为了确定控制过程在编码和提取过程中如何改变基于时间和意义的组织的平衡，我们的实验将为受试者提供一个目标，并改变与该目标最相关的关联类型。我们的模拟将测试这样的假设，即可以通过向网络添加目标表示层来解释目标的影响。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

Memory links our experiences together. Remembering one thing can remind of us another. This can be a blessing, as when a song transports you back in time to relive a fond memory. Or it can be a curse, as when seeing a colleague reminds you of a missed deadline. Controlling which memories come to mind influences one's well-being. This research will study why certain memories become linked and the processes that allow us to search for specific memories and avoid others. Existing work has often used tasks that simplify real-world memories by having participants only study random words. We will look at how adding real-world complexities to tasks impacts behavior. Another feature of our work is the use of computer models to simulate the cognitive mechanisms hypothesized to underlie memory search. These simulations allow us to evaluate theories of memory. This work will also help educate the next generation of scientists. We will train graduate students in our laboratory and develop a workshop for local high schools. Students will collect and analyze data. This will introduce them to both memory science and computer programing.The research has two scientific objectives: Objective 1 is to elucidate the computational basis of interactions between time-based and meaning-based associations. Events that occur nearby in time tend to be related by meaning, but typical free-recall tasks lack meaning-based relations between items. Therefore, subjects rely heavily on time-based associations to guide memory search: recalling a word and using it as a cue to recall another that was studied nearby in time. But recent data reveal that subjects all but abandon time-based associations when lists contain strong meaning-based associations. To ascertain how temporal and semantic information interact to guide memory search, our experiments will vary the number and nature of semantic associations in lists. Our simulations will test the hypothesis that such interactions can be explained by representing semantic similarity and temporal proximity in separate layers of a network. Objective 2 is to elucidate the computational principles enabling control processes to change the dynamics of memory encoding and search. We often form new memories without trying and later find that cues bring them spontaneously to mind. Other times, we deliberately encode information which we will need to accomplish some goal. But most laboratory memory tasks are not quite like these situations: subjects deliberately memorize items but only for the sake of rote memory and not for a specific goal. Recent data show that both incidental and goal-directed encoding markedly alter subjects' reliance on time-based associations. This suggests that control processes, which allow the memory system to adapt to different task demands, are critical determinants of how memory is organized. To ascertain how control processes change the balance of time- vs. meaning-based organization during both encoding and retrieval, our experiments will provide subjects with a goal and vary the type of association most relevant to that goal. Our simulations will test the hypothesis that the influence of goals can be accounted for by adding a goal representation layer to the network.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.