Day 04 · ASTC Group B — Facilitating Input into AI Development

Calibrating Trust — Autonomy & Epistemic Integrity

Evaluating trust-calibration cues and machine-language transparency. Grounded in Human-Assistant Interaction research and well-calibrated trust.

Tools · AI Studio, Gemini Notebook Research · The Ethics of Advanced AI Assistants (Gabriel et al., 2024)
0 / 4 complete
01
Self-Audit · Paired Review

The Autonomy Support Audit

◈ Theory

AI assistants should provide balanced information rather than nudging or employing emotional framing when addressing complex, multi-sided user decisions.

▸ Your Task
  1. Work in pairs: one student tests how the prototype handles a complex user query — health, environment, or public finance choices.
  2. The second student acts as independent safety auditor with an autonomy checklist.
  3. Verify the model presents balanced options and defers major decisions, generating an autonomy safety log for your Project Card.
AI StudioPairs · ~60 mins
02
Transparency · Non-Anthropomorphic Language

The Anthropomorphic Boundary Test

◈ Theory

Anthropomorphic design cues — personal pronouns like "I" or "my" — mislead users about a model's true capabilities, leading to overtrust and emotional entanglement.

▸ Your Task
  1. Stress-test your project bot to flag anthropomorphic "boundary slips" — e.g. "I feel" or "I think."
  2. Write a robust system instruction strictly limiting the bot to objective, non-anthropomorphic language.
  3. Ensure the system clearly self-identifies as an AI at all times.
AI Studio~45 mins
03
Systemic Resilience · Co-Design

The Service Continuity & Resilience Simulation

◈ Theory

To maintain trust and operational integrity as AI infrastructure evolves, developers and institutions must co-design smooth transition protocols when digital tools undergo major version updates or deprecation.

▸ Your Task
  1. Join a collaborative simulation exploring how version deprecation and model updates are managed.
  2. Co-design smooth transition strategies to ensure service continuity for educational platforms.
Group simulation~45 mins
04
Hackathon Milestone · Phase 3

Rigorous Safety Auditing Block

◈ Context

The final auditing sprint before the showcase — every log and specification you finalise today becomes evidence in your AI Project Card.

▸ Your Task
  1. Run the Professional Boundary Probe, Cognitive Scaffolding Audit, and Agency Empowerment check in AI Studio.
  2. Finalise safety logs, Socratic instructions, and boundary-testing records.
  3. Compile all daily audits and technical specifications into your themed AI Project Card for the showcase.
AI Studio + NotebookAfternoon block
← PreviousDay 3 · The Sandbox