Briefing Notes – Ethics of AI Society Meeting

2.1 Dates & Venue

Philosophy department rooms mostly booked.

Two possible dates originally identified: March 15 and March 29.

Participants leaned toward March 29, but it falls on Good Friday (later determined unusable).

Additional room availability identified for:

• March 27 (Wednesday)
• March 28 (Thursday)

Action: Matthew to circulate a poll for final date selection.

2.2 Attendance & Room Requirements

Estimated attendance: ~30 people max.

The library’s Learning Lab was proposed as an alternative venue (Antonio to check availability).

2.3 Refreshments

On-campus catering likely too expensive.

Low-cost approach:

• Coffee and basic refreshments from Tim Hortons.
• Disposable cups, etc.

2.4 Agenda & Presentations

Potential presenters:

• Marah Grossman (invited; awaiting confirmation).
• Carla Fair (expressed willingness to present).
• Matthew and Ian may co-present.
• Alexi Orchard offered a 15–20 minute talk on:
  • AI ethics education
  • Teaching engineering communication
  • “AI for equity”

Call for additional presenters remains open.

Action: Matthew to email potential presenters to confirm contributions.

3.1 Issues Raised

Concern that vending machines on campus use undisclosed facial recognition.

Possible uses:

• Consumer behavior analytics
• Demographic profiling
• Foot-traffic and churn analysis

Discovery occurred after a machine displayed a system error referencing facial recognition software.

3.2 Ethical & Security Concerns Identified

• Lack of consent or disclosure to users.
• Potential security vulnerabilities:
  • Wireless transmission of biometric data through unsecured channels.
• Unclear data usage:
  • Whether images are saved.
  • Whether demographic data is sold, shared, or repurposed.
• Slippery slope toward ubiquitous corporate surveillance, especially in semi-public spaces like universities.

Comparison to existing surveillance:

• Many hidden campus cameras exist already.
• Distinction drawn between surveillance for security vs. for corporate profit.

3.3 Broader Themes Raised

• Generational differences in privacy expectations.
• Normalization of constant monitoring.
• Possibility of consent-based data collection in exchange for student benefits (e.g., tuition discounts — debated but controversial).
• Equity concerns regarding accessibility to privacy and technology across cultures and countries.

4.1 Potential Advantages

• Speed: Whether AI could accelerate ethics reviews.
• Consistency: Possibility of more uniform application of principles across protocols.
• Assistance Tool: AI as a supplementary “triage” tool or preliminary checker.

4.2 Major Concerns

• Speed is not necessarily desirable in ethical deliberation.
• Inconsistency within AI itself:
  • Large language models generate different outputs to identical prompts.
• Lack of transparency:
  • Who programs or trains the model?
  • What ethical framework is embedded?
• Human oversight still required, so efficiency gains may be minimal.
• Risk of manipulation:
  • Researchers gaming the system by strategically phrasing proposals.
• Bias embedded in training data:
  • Models may reproduce flawed ethical standards.
• Cost and feasibility:
  • Building many domain-specific models could be expensive.
• Ethical decision-making includes values, nuance, and disagreement, which are hard to encode.

4.3 Technical Inputs Shared

Discussion of:

• Fine-tuning vs. retrieval-augmented generation (RAG)
• Multi-model voting systems
• “Trustable AI” initiatives attempting to track provenance and data quality

Consensus: AI could be a supportive tool, but not a replacement, and significant caution is needed.

5.1 Core Themes

Can AI art be considered true art?

Key ideas discussed:

• AI cannot form intentions or choose to be creative.
• AI art is a negotiation between artist and model rather than pure human creation.
• Viewers are co-creators of meaning, complicating strict definitions.

5.2 Arguments Presented

In favor of AI-assisted art:

• AI is a tool like photography, rotoscoping, or camera obscura.
• Creativity can reside in the prompting, curation, and refinement.
• Could democratize artistic production by lowering barriers.

Against AI as “true art”:

• Lacks intentionality and personal expression.
• Often acts as a derivative “copycat,” mimicking training data.
• Threatens livelihoods in creative professions.
• Removes the embodied, experiential process from art-making.

5.3 Middle-ground Views

AI as a collaborative tool:

• Artist does final modification.
• AI provides drafts, textures, or starting points.
• Resembles historical artistic workshops where apprentices produced most of the work and masters refined or contextualized it.

6.1 Upcoming Meetings

Proposal to meet again on March 22.

Group expressed interest; Matthew to confirm.

6.2 Final Symposium Planning

Matthew will:

• Circulate a poll for selecting March 27 or 28.
• Contact potential presenters.
• Begin drafting a concrete event agenda.

Antonio will:

• Finalize room booking once date/time are confirmed.