14 Adaptive Multi-Source Learning
How should an intelligent system decide whom and what to trust?
Area 14: Adaptive Multi-Source Learning
Master Question
How should an intelligent system decide whom and what to trust?
What We Want to Discover
An agent often receives information from several sources at once: sensors of different quality, other agents, human advisors, or prior models. This area studies how the agent should estimate the reliability of each source, how that estimate should change over time and with evidence, and how conflicting sources should be reconciled.
Why It Matters
Blind averaging of sources performs badly when some of them are wrong, stale, or adversarial. A system that can learn who or what to trust degrades more gracefully when part of its information supply fails or misleads it.
Core Concepts
- Source reliability estimation
- Trust modeling and calibration
- Multi-fidelity learning
- Conflict resolution between information sources
- Bayesian model averaging
Relevant Disciplines
- Epistemology
- Social psychology of trust
- Statistics
- Sociology of expertise
Potential Mimicry Sources
- Human trust formation and betrayal
- Reputation systems in animal and human societies
- Expert elicitation in forecasting
Projects
No projects yet.
Active Questions
No projects yet, so there are no active derived questions to report.
Key Findings Across Projects
Pending. No projects in this area have produced findings yet.
Unresolved Questions
Pending.
Connections to Other Areas
Pending. Connections will be identified as projects in this area develop.