Beyond Traditional Teaching: Large Language Models as Simulated Teaching Assistants in Computer Science
Bibliographic record
Abstract
As the prominence of Large Language Models (LLMs) grows in various sectors, their potential in education warrants exploration. In this study, we investigate the feasibility of employing GPT-3.5 from OpenAI, as an LLM teaching assistant (TA) or a virtual TA in computer science (CS) courses. The objective is to enhance the accessibility of CS education while maintaining academic integrity by refraining from providing direct solutions to current-semester assignments. Targeting Foundations of Programming (COMP202), an undergraduate course that introduces students to programming with Python, we have developed a virtual TA using the LangChain framework, known for integrating language models with diverse data sources and environments. The virtual TA assists students with their code and clarifies complex concepts. For homework questions, it is designed to guide students with hints rather than giving out direct solutions. We assessed its performance first through a qualitative evaluation, then a survey-based comparative analysis, using a mix of questions commonly asked on the COMP202 discussion board and questions created by the authors. Our preliminary results indicate that the virtual TA outperforms human TAs on clarity and engagement, matching them on accuracy when the question is non-assignment-specific, for which human TAs still proved more reliable. These findings suggest that while virtual TAs, leveraging the capabilities of LLMs, hold great promise towards making CS education experience more accessible and engaging, their optimal use necessitates human supervision. We conclude by identifying several directions that could be explored in future implementations.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.008 | 0.032 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.004 | 0.005 |
| Open science | 0.002 | 0.004 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.005 | 0.002 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".