AI Education Thesis Ideas

Key thesis directions from UCSD's AI integration:

• Evaluate Socratic AI tutors on learning outcomes in large classes: Measure engagement variations,...

Key empirical insights for LLM-based ITS in programming courses:

• Negative impact from heavy LLM use on demanding tasks hurts analytical skills and...

Core thesis direction: Build authentic simulated students for math teacher training using real data sources.

• Data foundation: 917 utterances from...

Prime thesis opportunities in AI tutoring systems that reinvent teaching:

• Function as adaptive cognitive interfaces analyzing student responses
  -...

Analysis Tools for Tutoring Data

• 🔥 Sandpiper App: New open-source app from National Tutoring Observatory uses agentic AI to analyze large...

Unlock scalable learning analytics with Sandpiper, NTO's open-source app using agentic AI to annotate tutoring transcripts at utterance level.

Key...

• Over-reliance risks: Chatbot tutors often spoon-feed solutions, hindering absorption and failing to beat traditional methods.
• Promising fix: UPenn...

A two-stage LLM method pinpoints student misconceptions in biomedical courses, ripe for thesis extension into intelligent tutoring systems:

• Analyzed...

Papers to Read

• 🔥 SKILL0: Paper on In-Context Agentic Reinforcement Learning for Skill Internalization.
• 🔥 AI-Based Automated Scoring Layer:...

SKILL0 explores in-context agentic reinforcement learning for skill internalization – ripe for edtech theses adapting RL to personalized skill acquisition and intelligent tutoring.

Massive online courses face navigation crisis:

• 400+ lectures (30-90 mins each) yield just 22% completion
• Students scrub for sections, then quit
  -...

Emerging framework for automated scoring of open-ended questions unlocks thesis potential:

• Core method: LLMs + RAG + rubrics enable criterion-based,...

Emerging hybrid framework for EFL writing assessment offers rich thesis potential in AI-edtech:

• Integrates process/product evaluation via tech...

Thesis idea: Adapt DCR to prioritize hard MCQs in intelligent tutoring systems.

• Divides questions by confidence score from answer frequencies.
  -...

Automated Short Answer Grading

• 🔥 Negation-Sensitive Calibration: Negation-Sensitive Calibration for Embedding-Based Automatic Short Answer...

• Core strengths: Improves language accuracy, expands vocabulary, supports revision
• Notable limits: Weaker impact on other areas
• Thesis potential: Tackle open questions in personalization and L2 revision support

Key thesis direction from recent work on ASAG:

• Negation-sensitive calibration targets embedding-based grading of short free-text responses
• Aims...

• Attack Framework for Ed Prompts: Decompose learning prompts into segments, build role-consistent attacks, and embed stealthy payloads in student...

Key thesis directions from HA-BCHM advancements in automated essay scoring:

• Linguistic feature engineering: Fuse CNN for local cues (grammar,...

How can PERMA's event-driven framework advance thesis work on personalized learner modeling in intelligent tutoring systems?

• Dynamic interactions:...