Reproducibility as a Mechanism for Teaching Fairness, Accountability, Confidentiality, and Transparency in Artificial Intelligence
Project Overview
The document highlights the integration of generative AI in education through a graduate-level course titled Fairness, Accountability, Confidentiality, and Transparency in Artificial Intelligence (FACT-AI) at the University of Amsterdam. This course is designed to engage students in ethical AI practices by emphasizing reproducibility, where they work on group projects to replicate existing FACT-AI algorithms, culminating in submissions to the Machine Learning Reproducibility Challenge. Key applications of this approach include enhancing students' understanding of ethical considerations in AI and developing practical skills in algorithm reproduction. The findings indicate that the course has led to high levels of student satisfaction, with participants appreciating the hands-on experience and the relevance of the course content. Additionally, the successful submissions to academic challenges demonstrate the effectiveness of this pedagogical method in fostering a deeper comprehension of generative AI's implications in real-world scenarios. Overall, the course serves as a model for incorporating ethical AI discussions and practical applications into higher education, preparing students to navigate the complexities of AI technology responsibly.
Key Applications
FACT-AI course focused on reproducibility
Context: Graduate-level course for MSc AI students at the University of Amsterdam
Implementation: Students work in groups to reproduce existing FACT-AI algorithms from top AI conferences and submit reports to the Machine Learning Reproducibility Challenge.
Outcomes: Students gained hands-on experience with reproducibility, contributing to the open-source community and publishing reports in academic venues.
Challenges: Challenges included adapting the course to an online format during the COVID-19 pandemic and ensuring group dynamics in remote settings.
Implementation Barriers
Educational Barriers
Students struggled with group dynamics and communication in an online format due to the pandemic.
Proposed Solutions: Utilizing various communication tools (WhatsApp, Discord, Slack) and maintaining regular contact with teaching assistants.
Logistical Barriers
Some students faced challenges in understanding the requirements for formal submission of their reproducibility reports.
Proposed Solutions: Providing explicit instructions on submission processes and examples of high-quality reproducibility papers.
Project Team
Ana Lucic
Researcher
Maurits Bleeker
Researcher
Sami Jullien
Researcher
Samarth Bhargav
Researcher
Maarten de Rijke
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Ana Lucic, Maurits Bleeker, Sami Jullien, Samarth Bhargav, Maarten de Rijke
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai