Human-AI Co-Mentorship in Project-Based Learning: A Case Study in Financial Forecasting
Human-AI Co-Mentorship in Project-Based Learning: A Case Study in Financial Forecasting
Authors: Freyaa Chawla, Ahan Chawla, Rishi Singh, Joe Germino, Grigorii Khvatskii Date: 2026-05-06 Paper ID: openalex:2605.05144
Summary
This paper presents a pedagogical case study on using AI tools to facilitate research projects for students with limited prior technical experience. By emphasizing a workflow-first approach, the project enabled high-school participants to focus on high-level problem formulation and financial analysis while offloading technical implementation to AI assistants. The study demonstrates that such co-mentorship models effectively bridge skill gaps and accelerate learning in complex domains like financial forecasting.
Key Contributions
- Demonstrates a pedagogical framework where students leverage AI tools to execute technical workflows, allowing focus on high-level problem formulation rather than prerequisite mastery.
- Reports on a case study where students with limited technical backgrounds successfully developed functional financial models for ETF price prediction.
- Introduces a ‘workflow-first’ pedagogical model that integrates daily human-to-human stand-ups for conceptual debugging and AI-assisted iterative coding.
Open Questions & Future Work
Archivist Review
The paper focuses on a pedagogical case study rather than a technical innovation in time-series forecasting. The proposed pedagogical framework was rejected as it is outside the scope of the ML architecture and methodology vault. The open question regarding LLM sentiment validation in finance was approved as it touches on a relevant, unresolved limitation in financial modeling performance.
Approved Open Questions
- Validating LLM Sentiment Metrics: The study identified the lack of robustness in LLM-generated sentiment scores as a potential reason for performance degradation, highlighting the need for rigorous validation methods to determine the actual utility of sentiment in financial modeling.
Rejected Candidates
- [concept] Workflow-first pedagogical model (
workflow-first-pedagogical-model) - not_reusable: This is a pedagogical framework for student mentoring and lacks direct relevance to the time-series forecasting methodology or model architecture core to the research vault.
Links
Metadata & Links
- url
- https://arxiv.org/abs/2605.05144
- paper_id
- 2605.05144
- paper_source
- openalex
- domain
- nlp
- tags
-
- architectures
-
- datasets
-
- concept_slugs
-
- dataset_slugs
-
- skill
- TimeSeriesSkill
- created_at
- 2026-05-09T05:57:47Z