What is AI Hallucination?
AI hallucination refers to instances where a language model generates content that appears plausible but is actually incorrect, fabricated, or not supported by its training data or provided context. The term draws an analogy to human hallucinations - perceiving things that aren't there.
Types of Hallucinations
Factual Errors
- Incorrect dates, names, statistics
- Non-existent citations
- Made-up historical events
Logical Inconsistencies
- Self-contradicting statements
- Invalid reasoning chains
- Impossible scenarios
Confabulation
- Filling gaps with plausible fiction
- Inventing details
- Mixing up related concepts
Why Hallucinations Occur
- Training data limitations
- Probabilistic nature of generation
- Lack of real-time knowledge
- No inherent fact-checking
- Optimization for fluency over accuracy
Mitigation Strategies
Technical
- Retrieval-Augmented Generation (RAG)
- Grounding with verified sources
- Uncertainty quantification
- Chain-of-thought prompting
Operational
- Human review for critical outputs
- Citation requirements
- Confidence thresholds
- Domain-specific fine-tuning
Detection Methods
- Cross-reference with known facts
- Consistency checking
- Source verification
- Uncertainty estimation