Evolution of AI in Education
Evolution of AI: Symbolic to Deep Learning
- Symbolic Systems (1950s–1980s)
- Rule-based reasoning
- Logic, if-then rules
- Limited scalability
Arthur Samuels
The Turing Test: A Philosophical Starting Point
Can a machine’s behavior be indistinguishable from that of a human?
Alan Turing (1950) posed the question:
“Can machines think?”Proposed the Imitation Game, now known as the Turing Test
A machine passes if a human evaluator cannot reliably distinguish it from a human in conversation
- Still debated today:
- Does passing the test mean a machine is intelligent?
- Is mimicking enough in educational contexts, or do we need true understanding?
Live Turing Test
Machine Learning (1990s–2010s)
- Pattern recognition from data
- Requires labeled examples
- Algorithms “learn” from experience
Deep Learning (2010s–Present)
- Neural networks & large-scale computation
- Enables language, image, and speech processing
- Powering today's generative AI (e.g., ChatGPT, image
Retina detects raw light
Input layer gets raw pixels
“These pixels are dark, those are light”
V1 neurons detect edges and orientations
First NN layer learns edge detectors
“This is a vertical line”
V2/V4 neurons detect shapes & textures
Middle NN layers detect patterns
“This looks like fur”
Inferotemporal cortex recognizes objects
Final NN layers classify categories
“That’s a cat”
AI Capabilities in Education
Think-Pair-Share 🤔
What are the possible opportunities?
What are the limitations?
Debrief
Capabilities
- Adaptive learning platforms
- Automated grading (especially objective items)
- Intelligent tutoring systems
- Content generation
- Predictive analytics for at-risk students
Limitations
- Understanding nuance in student reasoning
- Supporting emotional/social learning
- Generalizing across educational contexts
- Equity and bias concerns
- Often “black box” systems with limited transparency
Frameworks for Evaluating AI Tools
Pedagogical Fit
Data Usage
Implementation Needs
Frameworks for Evaluating AI Tools
Pedagogical Fit
- Does it align with constructivist, behaviorist, or other pedagogical models?
- Is it enhancing or replacing human instruction?
Data Usage
- What data is being collected and why?
- Is it FERPA-compliant?
- Are students and educators informed?
Implementation Needs
- Infrastructure requirements
- Teacher training and support
- Integration with existing LMS or platforms
- Cost and scalability
Educational Problems AI Can Help Solve
- Scaling personalized feedback
- Detecting disengagement early
- Supporting multilingual or neurodiverse learners
- Automating routine administrative tasks
Problems AI Struggles With
- Fostering authentic student motivation
- Teaching critical thinking and creativity
- Supporting complex social-emotional learning
- Adapting to rapidly changing classroom dynamics