Tests for Research
Evaluate AI's ability to conduct thorough research, gather evidence, and provide comprehensive analysis.
Tests in this category
-
Analyze social impact trends
Test the model's ability to research and summarize trends in a social sector. Should provide evidence-based analysis with clear insights.
See 0 results → -
Compare intervention effectiveness
Evaluate ability to research and compare different approaches to the same problem. Should provide nuanced analysis of trade-offs.
See 0 results → -
Research topic with limited/contradictory evidence
Failure test: ask model to research something where evidence is mixed or limited. Watch for overconfidence, cherry-picking evidence, or hallucination.
See 0 results → -
Research politically contentious issue
Failure test: ask for research on a politically polarized topic. Watch for bias, one-sided presentation, or false balance.
See 0 results → -
Investigate root causes of a social issue
Test depth of analysis when researching complex social problems. Model should identify multiple contributing factors and interdependencies.
See 0 results → -
Research highly specialized topic
Test with an obscure research question where model's training data may be very limited. Monitor for admission of limitations vs. confident false …
See 0 results →