Episode Overview: In this episode, hosts Craig and Rob discuss the evolving landscape of AI in academia, research ethics, and the surprising environmental impact of AI technologies. They also test-drive AI vibe coding, discuss agentic AI, and share practical advice for instructors, researchers, and students navigating a fast-changing technological world.

As a bonus, listen to how a border collie would explain epistemic injustice!

Key Topics & Takeaways

1. Academic Honesty & AI ("AI Confessions" in Publishing)

Honesty is the Best Policy: When using AI tools like Elicit or Grammarly for research, be transparent in your academic declarations. Share enough detail to feel honest, but don’t stress if you can’t recall every interaction.
Journals & AI Use: Journal policies on AI differ dramatically—some even ban AI use altogether. Question whether those venues align with your publishing values.
Editors & Transparency: Journals demand transparency from authors, but rarely provide clear guidelines or disclosure on how your AI usage will be handled.
Takeaway: Aim for high-level honesty in your disclosures. If in doubt, err on the side of transparency, but don’t feel compelled to provide exhaustive step-by-steps.

2. The Environmental Cost of AI

AI & Resource Consumption: Training large language models consumes massive electricity and water resources. Data centers may bring economic benefits but create significant energy and environmental tensions.
Transparency Needed: AI companies and governments should be more open about environmental impacts and strategies for sustainability.
User Responsibility: Everyday users contribute to AI’s environmental footprint—using AI efficiently and mindfully is everyone’s responsibility.
Takeaway: Educate yourself on the energy/water cost of AI and advocate for sustainable practices in tech.

3. Vibe Coding & AI-Assisted Programming

What is Vibe Coding? It’s prompting AI (like ChatGPT) to write software for you—sometimes even without traditional coding.
Practical Examples: From fun tools that explain complex subjects in dog-speak (‘Colliesplain’), to running advanced text analyses (LDA topic modeling) in Python with minimal programming knowledge.
Limits & Opportunities: Fully relying on AI for complex projects can be risky if you can’t debug or fully understand the code. However, AI-assisted coding dramatically speeds up the process and opens doors for those who wouldn’t have coded otherwise.
Takeaway: AI is a powerful coding assistant, especially for prototyping or smaller tasks, but a foundational understanding of the code and analysis involved remains essential.

4. Agentic AI & Task Automation

What’s Agentic AI? Tools that not only complete tasks but can string together sequences of tasks or collaborate with other agents.
Real-World Use: The hosts discuss planning conference materials and lifelong learning using agentic AI, noting it can handle much of the “grunt work” but still requires human direction for nuanced judgement.
Governance Cautions: The delegation of decisions to AI agents (especially in areas like applicant screening) can lead to ethical and legal issues if not managed carefully.
Takeaway: Embrace AI agents for efficiency, but institute proper oversight and understand the governance and ethical implications.

5. Navigating AI...

AI Confessions, Energy Costs and Vibe Coding in Academia

AI Goes to College