Previous Meetings

Briefing Notes – Ethics of AI Society Meeting

• Explore ethical, practical, and governance issues surrounding AI.
• Discuss implications for academia, research, and society.
• Plan next meeting and potential symposium.

Next Meeting:

• Date: November 16, 2023, at 3:00 PM.
• Location: Same meeting room as current session.

Potential Symposium:

• Consider a one-day event with invited speakers in November or later.

Action:

• Matthew to confirm room booking and share symposium updates.
• Participants to propose topics for next meeting via Teams.

A. Introductions & Interests

• Participants include faculty, librarians, teaching support staff, data scientists, and students.
• Key interests:
  • Ethics of belief and responsibility in AI use.
  • Data governance and bias in datasets.
  • Hidden labor in AI systems (e.g., content moderation).
  • Impact on teaching and research integrity.

B. Natural vs. Artificial Intelligence

• Example: Bees trained to detect explosives combined with AI monitoring.
• Concern: Over-reliance on AI may overshadow natural intelligence and human judgment.

C. Hidden Labor & Data Labeling

• Workers labeling harmful content face severe mental health risks.
• Outsourcing to low-cost labor markets raises ethical concerns.
• Some workers use AI to automate labeling, risking data quality.

D. Data Poisoning

• Technique: Altering pixels invisibly to mislead ML models.
• Risks:
  • Protecting intellectual property vs. malicious sabotage.
  • Misuse in critical domains (e.g., medical imaging).
• Ethical dilemma: Potential market for “poisoning services.”

E. Governance & Informed Consent

• AI development outpaces policy.
• Academic concerns:
  • Students required to submit work to plagiarism tools (e.g., Turnitin).
  • Limited transparency on student data usage.
• Broader question: Who ensures accountability and consent in data use?

F. AI in Peer Review & Grading

Pros:

• Speed and scalability.
• Potential diversity of feedback.

Cons:

• Algorithms cannot evaluate originality.
• Risk of bias and opacity.

Suggested Role: AI as supplementary tool; human editors retain oversight.

G. Broader Ethical Concerns

• Opacity & Accountability: Hidden biases may evade detection.
• Manipulation Risks: Systematic influence at scale.
• Cultural Impact: Generative AI blurs authenticity (e.g., AI-generated Oasis album).
• Historical Parallel: Compared to nuclear technology—difficult to contain once developed.

• Transparency & Explainability: Essential for trust.
• Human Oversight: Critical in high-stakes domains.
• Regulatory Lag: Voluntary codes insufficient.
• Ethical Responsibility: Developers must anticipate societal impact.

• Confirm November 16 logistics.
• Share symposium details and invite topic suggestions.
• Continue exploring governance frameworks and safeguards.