Abstract
This study explores the impact of non-cognitive factors—specifically social skills, personal goals, and institutional integration—on the college graduation rates of deaf and hard-of-hearing students. Using an Agent-Based Model (ABM) grounded in Tinto’s Model of Student Retention, the researchers simulated individual student-teacher interactions over a four-year period. Results indicate that while academic performance is a necessary baseline, social integration and clear goal-setting are the primary drivers of long-term persistence, particularly during the critical first year. Sensitivity analysis reveals that improvements in academic experience yield the most significant gains in retention.
Introduction
While post-secondary enrollment for deaf and hard-of-hearing students has increased significantly since the 1980s, their graduation rates remain disproportionately lower than those of their hearing peers. Traditional educational research has historically focused on cognitive factors, such as literacy levels or GPA, to predict academic success. However, empirical data suggests that even academically prepared deaf students frequently depart from higher education before completion.
This paper addresses the "departure decision" by examining non-cognitive factors—the social and psychological elements of the college experience. Understanding how these invisible factors interact over time is challenging for traditional statistical methods, which often fail to capture dynamic, individual-level interactions. Consequently, the authors employ computational modeling to bridge this gap and provide a more granular view of student persistence.
Methodology
The researchers developed an Agent-Based Model (ABM) using the NetLogo software environment. ABM offers a "bottom-up" approach, simulating the autonomous actions and interactions of agents (Deaf Students and Teachers) to observe emerging macro-level patterns.
Theoretical Framework
The model is grounded in Tinto’s Model of Student Retention, a seminal sociological theory that posits persistence as a result of a student's successful integration into both the academic and social systems of an institution.
To make the simulation computationally tractable, the researchers developed a simplified version of this framework, focusing on four key non-cognitive variables: Goal, Academic Experience, Social Skill, and Social Integration.
Simulation Design
The simulation initializes 200 student agents. Each "tick" represents one academic year. During the "College" phase, students randomly form links with teachers and peers, which dynamically updates their social and academic integration scores. These scores, along with initial parameters derived from National Deaf Center (NDC) reports and existing longitudinal literature, determine whether a student persists or departs.
Results and Key Insights
The simulation revealed several critical trends regarding student retention:
- The First-Year Criticality: The majority of student departures occur during the first year (the first "tick"). Students who persist beyond this initial period show a significantly higher probability of eventual graduation.
- Sensitivity to Academic Experience: Sensitivity analysis identified Academic Experience as the most influential factor. Incremental improvements in this area resulted in larger gains in overall persistence compared to other variables.
- The Persistence Ceiling: Even with non-cognitive factors optimized to their maximum levels, the departure rate never reached zero. This suggests that external factors not included in the model, such as financial constraints or health issues, continue to play a role in student outcomes.
- Optimal Persistence Parameters: Using the "Behavior Search" tool, the highest graduation rate (fitness score of 88.7) was achieved when Goal, Social Skill, and Social Integration were maximized (1.0) and Academic Integration was at 0.8.
Discussion and Conclusion
The findings underscore that social skills and clear goal-setting act as vital buffers, particularly in the later years of a degree program. However, the study also highlights the limitations of current modeling, including the exclusion of financial and family background variables.
Future research should aim to couple ABM with Machine Learning to create adaptive agents capable of "learning" from their environment. Such advancements would provide policy-makers in deaf education with more accurate predictive tools to design effective intervention strategies.
Summary updated