Lecture 6 - AI as a social actor
| ← Back to AICS | Next: Lecture 7 → |
Lecture 6 discusses the rise of AI systems with socioaffective design features. We consider the growth of anthropomorphic and relatability-seeking AI, including companion applications. We discuss the impact of excessive validation (sycophancy), how design features of these products may prolong engagement and maximise self-disclosure, and consider evidence that they have a negative effect on wellbeing. We ask what ‘AI psychosis’ may be and how it is provoked. We consider the impact of AI relationships on people, and discuss the outlook for the future, including the possibility that AI systems are considered moral patients in need of “welfare”.
What you need to understand:
- The intentional Stance
- Steering vectors
- Companions applications
- AI Sycophancy
- AI psychosis
- “wanting” vs. “liking”
- AI Model welfare
Sample essay questions
What psychological impact do AI models with socioaffective design features have on their users?
AI systems are increasingly being judged to be “friend” rather than “tool”. What implications does this have for society?