🤖 AI

Profiles of AI Dependency: A Latent Class Analysis of Filipino Students' Academic Competencies

This study utilizes Latent Class Analysis on 651 Filipino college students in Pampanga to identify four distinct AI dependency profiles, revealing that students with high AI reliance exhibit significantly weaker academic competencies, thereby underscoring the urgent need for educational policies that balance AI integration with the preservation of critical thinking and essential academic skills.

Original authors: Emerson Q. Fernando, Julius Ceazar G. Tolentino, Maria Anna D. Cruz, Jordan L. Salenga, Vernon Grace M. Maniago, Juvy C. Grume, Erika M. Pineda, Aileen P. De Leon, John Paul P. Miranda

Published 2026-05-01

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Emerson Q. Fernando, Julius Ceazar G. Tolentino, Maria Anna D. Cruz, Jordan L. Salenga, Vernon Grace M. Maniago, Juvy C. Grume, Erika M. Pineda, Aileen P. De Leon, John Paul P. Miranda

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a group of students as a team of explorers trying to solve a maze. In the past, they had to use their own brains, maps, and compasses to find the way. Today, they have a powerful new tool: a "Magic GPS" (Artificial Intelligence) that can instantly tell them the shortest path.

This paper is a study of how Filipino college students are using this Magic GPS. The researchers wanted to know: Are these students still learning how to navigate the maze themselves, or are they letting the GPS do all the work for them?

Here is the breakdown of their findings in simple terms:

The Setup: Who and Where?

The researchers looked at 651 college students in the province of Pampanga, Philippines. They asked these students about how they use AI for five specific tasks:

Critical Thinking: Solving problems on their own.
Writing: Putting words on paper.
Independence: Learning without help.
Research: Finding facts and sources.
Engagement: Staying interested in class.

The Big Discovery: Four Types of Explorers

Using a special statistical method (think of it as a "sorting machine" that groups people by their habits), the researchers found that students aren't all the same. They fall into four distinct groups:

The "Super Navigators" (Highly Engaged Independent Learners):
- Who they are: These students have the strongest skills. They use the Magic GPS only when they really need a quick check, but they still know how to read the map themselves.
- Result: They are the most skilled at thinking, writing, and researching.
The "Smart Co-Pilots" (Selective AI Users):
- Who they are: These students are good at writing and researching. They treat the GPS like a helpful assistant—using it to speed things up, but not letting it drive the car.
- Result: They have strong skills and use AI as a tool, not a crutch.
The "Drifting Drivers" (Moderate AI Users):
- Who they are: These students are okay at most things, but they are starting to rely on the GPS a bit too much, especially when it comes to learning on their own.
- Result: Their skills are average, but they are becoming a little too dependent on the tool.
The "GPS Zombies" (AI-Dependent Learners):
- Who they are: This group is in trouble. They have let the Magic GPS take the wheel completely. They barely think for themselves, write little on their own, and struggle to find information without the AI.
- Result: They have the weakest academic skills of all. They are so used to the AI doing the work that their own "muscles" (skills) are getting weak.

What the Data Tells Us

The Trend: Many students are using AI to do the heavy lifting for research and writing. While this makes things faster, the study found that the more they rely on it, the weaker their own fundamental skills become.
The "Senior" Problem: Interestingly, the older students (seniors) were more dependent on AI than the freshmen. It's like the longer they use the GPS, the less they trust their own ability to navigate.
No Gender or Grade Bias: It didn't matter if the student was a boy or a girl, or if they were a straight-A student or a struggling one. Both high achievers and lower achievers used AI in similar ways. The difference was how they used it (as a tool vs. as a replacement).

The Main Takeaway

The paper warns that while AI is a convenient shortcut, relying on it too much is like letting a robot do your homework. If you do that, you might get the answer, but you won't learn how to solve the problem yourself.

The researchers suggest that schools need to teach students how to use the GPS wisely (AI literacy) rather than just banning it or letting them use it blindly. The goal is to keep the students' own "navigation skills" (critical thinking and writing) strong, even while using the new technology.

In short: AI is a great tool, but if you let it drive the car entirely, you might forget how to drive. The study found that a specific group of students has already forgotten how to drive, and schools need to help them get back behind the wheel.

1. Problem Statement

The study addresses the growing concern that reliance on Artificial Intelligence (AI) tools, particularly generative AI (e.g., ChatGPT), is eroding fundamental academic competencies among higher education students. While AI offers benefits like real-time feedback and content generation, there is a risk of "cognitive offloading," leading to:

Reduced critical thinking and writing skills.
Diminished learning independence and research capabilities.
A shift from deep learning to superficial task completion.

Existing literature often focuses on general attitudes or isolated academic domains and lacks person-centered statistical approaches to identify behavioral profiles, especially in underrepresented regions like the Philippines. This study aims to fill that gap by analyzing AI dependency across five specific domains within the Philippine context.

2. Methodology

Research Design: Cross-sectional quantitative study.
Participants:
- Sample Size: 651 valid responses (out of 800 collected) from college students.
- Demographics: Students from 45 Commission on Higher Education (CHED)-accredited public and private institutions in Pampanga, Philippines.
- Profile: Mean age of 20.22 years; predominantly 1st and 3rd-year students; mixed gender distribution.
Instrument: A 24-item survey measuring five dimensions of academic skills:
1. Critical Thinking (CT)
2. Writing Skills (WS)
3. Learning Independence (LI)
4. Research Skills (RS)
5. Academic Engagement (AE)
- Reliability: High internal consistency confirmed via Cronbach's alpha ( $\alpha > 0.92$ for all constructs).
Data Collection: Online survey (Google Forms) conducted from October 2024 to January 2025 using snowball sampling.
Statistical Analysis:
- Descriptive Statistics: To analyze means and standard deviations of dependency scores.
- Latent Class Analysis (LCA): Used to identify distinct subgroups (profiles) of students based on their AI dependency patterns. Model fit was determined using Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC).
- Chi-Square Analysis: To examine relationships between demographic variables (gender, academic year, performance) and the identified latent classes.

3. Key Contributions

Person-Centered Profiling: Unlike previous studies focusing on variable-centered approaches, this study uses LCA to categorize students into distinct behavioral profiles regarding AI use.
Contextual Insight: Provides empirical data on AI dependency within the Philippine higher education system, a region often underrepresented in global AI-in-education literature.
Multidimensional Assessment: Evaluates AI impact across a holistic spectrum of academic skills (critical thinking, writing, independence, research, and engagement) rather than isolated tasks.
Policy Framework: Offers a basis for targeted educational interventions based on specific student profiles rather than a "one-size-fits-all" approach.

4. Results

Descriptive Findings:

Students showed moderate to high dependency on AI, particularly in uncritical acceptance of outputs ( $M=3.23$ ), research suggestions ( $M=3.13$ ), and substitution of academic engagement ( $M=3.06$ ).
Dependency was systematic rather than sporadic, indicating a broad erosion of core competencies.

Latent Class Analysis (LCA) Profiles:
The optimal model identified four distinct student profiles:

Highly Engaged Independent Learners (Class 3):
- Characteristics: Strongest competencies in all five domains; minimal AI dependency; use AI selectively.
- Status: Ideal academic performers.
Selective AI Users (Class 2):
- Characteristics: Strong research and writing skills; high engagement; use AI as a complementary tool rather than a primary resource.
- Status: Balanced and strategic users.
Moderate AI Users (Class 0):
- Characteristics: Moderate skills across domains but showing signs of dependency, particularly in learning independence.
- Status: At risk of slipping into over-reliance; need verification training.
AI-Dependent Learners (Class 1):
- Characteristics: Weakest academic competencies across all domains (especially CT, writing, and research); very high AI dependency; low learning independence and engagement.
- Status: High risk of academic skill erosion; likely substituting core skills with AI-generated content.

Demographic Correlations:

Gender: No significant difference in AI dependency patterns ( $p = .237$ ).
Academic Performance: No significant association between self-reported grades and dependency levels. Both high and low achievers utilize AI similarly, though high achievers may use it strategically while low achievers may rely on it for compensation.
Academic Year: Significant variation found ( $p = .003$ ). Senior students exhibited higher levels of AI dependency, likely due to increased academic workload and prolonged exposure to AI tools.

5. Significance and Implications

Educational Policy: Highlights the urgent need for AI literacy programs integrated early in the curriculum. Policies must balance technological adoption with the preservation of essential cognitive skills.
Targeted Interventions:
- AI-Dependent Learners: Require focused support to rebuild critical thinking and research skills.
- Moderate Users: Benefit from training in information validation and analytical reasoning.
- Independent Learners: Can serve as peer mentors to model balanced AI usage.
Curriculum Adaptation: Institutions must redesign assessments to prioritize critical thinking and process over final outputs, ensuring AI acts as a support tool rather than a substitute for learning.
Future Research: The authors call for longitudinal studies to track the long-term impacts of AI dependency and to develop frameworks for mitigating skill erosion in a technology-driven educational landscape.