Understanding Pain Points of ACT-R Modelers: A Human-Centered Approach

Restricted (Penn State Only)
- Author:
- Wang, Shan
- Graduate Program:
- Informatics
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- August 16, 2024
- Committee Members:
- Carleen Maitland, Program Head/Chair
Farnaz Tehranchi, Outside Unit & Field Member
Frank Ritter, Chair & Dissertation Advisor
Luke Zhang, Major Field Member
Aiping Xiong, Major Field Member - Keywords:
- Cognitive Modeling
Human-Centered Approach
ACT-R
Pain Points
Human-Computer Interaction
Error Modeling
Individual Differences
Cognitive Architectures - Abstract:
- Among the approaches in artificial intelligence, a classical one is to understand the human mind and, therefore, simulate human cognition and behavior. ACT-R, an example of a cognitive architecture, is a unified theory of cognition realized as a computer program. However, it is relatively difficult to learn, and its associated tools are not only challenging to use but also have limited applicability in various contexts. Additionally, concerns beyond the usability of ACT-R tools appear to exist within the broader research community. To gain a comprehensive understanding of these challenges, semi-structured interviews were conducted with ACT-R modelers (N=15) to explore their experiences with ACT-R. Specifically, the focus was on examining their pain points, the concerns of the broader research community, and their suggested solutions to the issues they encountered. As an example of an approach to simplify model-building, TAKLML (Task Analysis with Keystroke-Level Modeling and Learning) was introduced and applied. Two example applications of this approach to complex tasks are presented, demonstrating how it not only facilitates cognitive model building but also addresses challenging tasks within the field. In general, this dissertation contributes to the ACT-R community, cognitive architecture users, the HCI community, and the education community. It provides a relatively comprehensive framework with details for understanding the pain points in building ACT-R models, which can be used to enhance the usability of such models. Additionally, a potential partial solution to these challenges is proposed. This dissertation encourages further discussion regarding the gap between academic research and practical applications within the fields of Human-Computer Interaction and Human Factors. It also contributes to the field of education by emphasizing the importance of respecting diversity and individual differences with an open and growth mindset, so that educators do not “correct” students when they are applying different approaches, whether in problem-solving or learning.