Metacognitive Awareness Inventory (MAI): 52-Item Survey for LLM Self-Awareness

The Metacognitive Awareness Inventory (MAI) for LLM is a structured dataset based on the classic 52-item psychological instrument. It has been translated and adapted to evaluate self-awareness in large language models. The dataset is hosted on Kaggle, but the author, organization, and last update date are unknown.

Use Cases

• Benchmarking LLM self-awareness based on structured survey responses.
• Analyzing correlations between different metacognitive components as defined by the MAI framework.
• Training or fine-tuning models to exhibit or recognize metacognitive behaviors.
• Conducting comparative studies between human and LLM metacognitive profiles.

Strengths

• Based on the established 52-item Metacognitive Awareness Inventory, a known psychological instrument.
• Specifically structured and translated for a novel application in LLM evaluation.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Kaggle

Collection Method

Likely a structured translation and adaptation of the original MAI survey instrument.