Overview
What if you could simulate your leadership team's response to a crisis before it happens? Digital twins of decision-makers could enable organizations to stress-test strategies, identify blind spots, and practice responses to novel scenarios.
This research explored the technical and ethical dimensions of creating AI replicas that model individual decision-making patterns.
Why Paused
This project is currently paused due to unresolved questions around:
- Consent and representation: When does a digital twin become identity theft?
- Accuracy vs. caricature: Simple models risk reducing people to stereotypes
- Organizational dynamics: How teams react to "simulated colleagues" in practice
- Data sensitivity: The training data required raises significant privacy concerns
Initial Experiments
Before pausing, early experiments demonstrated:
- LLMs can capture surface-level communication patterns from meeting transcripts
- Decision-making style is harder to replicate than communication style
- Participants found "talking to themselves" unsettling in unexpected ways
- The most value came from the reflection process, not the twin itself
"The digital twin was wrong about what I'd do—but thinking about why it was wrong taught me something."
What Would Resume This
This research will resume if/when:
- Clearer ethical frameworks emerge for synthetic identity creation
- Organizations develop governance for digital twin consent and use
- Technical capabilities improve for modeling decision-making (not just language)