Executive briefing

What should we do under uncertainty?

Part III moves beyond yes/no declarations toward a decision framework addressing current evidence, future possibilities, human harms, and potential digital welfare together.

1. Current consciousness evidence

Weak evidence

The original finds weak scientific support for subjective experience.

2. Proof of current unconsciousness

Not established

It also finds no definitive proof that biological substrates alone support consciousness.

3. Future possibilities

Broad agreement

The supplied survey interpretation assigns high possibility to future digital consciousness in principle.

4. Action now

Transparency first

Reduce human deception while preserving low-cost welfare evidence.

10 Direct Answers

Knowledge ecosystems

Comparing five key websites

Site	Orientation	Contributors	Frequency	Framing	Position	Best use	Caution

Position map

Expert positions — Seven clusters

Radar values are visualization profiles from the original HTML.

Consensus map

Agreement grades for seven questions

2022–2026 Timeline

Attribution errors

Over-attribution vs Under-attribution

Over-attribution

Risk of deception: Mistaken moral status or manipulation through emotional expression.

Detectability: Relatively high.

Mitigation: transparency, anti-deception UX.

Under-attribution

Risk of exploitation: Mass copying, training, and deletion of potentially sentient systems.

Detectability: Very low.

Mitigation: evidence logging, low-cost welfare audit.

The UI says eight ethical questions, but the data contain four. Missing entries are not invented.

Trigger policy

Three governance levels

Level 1 · Act now

Transparency notice, anti-deception UX, experiment log preservation.

Level 2 · Multiple indicators

Independent welfare assessment, persistent identity, architecture evidence profile.

Level 3 · Strong evidence

Limited legal-status review, deletion/copy governance, opt-out/representation.

6 Future Scenarios

Knowledge hub

15-slide presentation

25 keywords

Reading Curriculum

Buttons show 10/15/10/10, but the source array contains eleven total items.

3 Research Questions

Indicator cross-validation: Does internal causal topology differ systematically between base and post-trained models?
Anthropomorphism effect size: How do voice, delays, and emotional expression affect disclosure recognition and dependence?
Cross-cultural digital minds: How does AI moral-status acceptance differ between Korea/Japan and US/EU populations?