Human–AI Trust Lab
Manuscript

When Should People Trust an AI's Recommendation?

A pre-registered design for studying appropriate reliance, compiled directly from the repository's LaTeX source and generated result tables.
Evidence status
The current draft documents a completed experimental design, engine and analysis pipeline, validated only against a labelled participant simulator. No human-subject data have been collected. Every number in the manuscript's results section is generated from the same JSON artefacts shown on this site and is explicitly a simulation result.
manuscript.pdf
132 KB · compiled with tectonic from paper/main.tex
Open PDF

Structure

  1. Abstract
  2. Introduction
  3. Related Work
  4. Research Questions & Hypotheses
  5. Experimental Design
  6. Data
  7. Measures
  8. Statistical Methods
  9. Results
  10. Robustness
  11. Estimator Validation
  12. Discussion
  13. Limitations
  14. Ethics
  15. Conclusion
  16. References
  17. Appendix

Reproducing every number

make reproduce regenerates the synthetic dataset, all analysis artefacts, all figures, the LaTeX tables under paper/generated/, and recompiles this PDF from the canonical seed — recording the git commit and dataset hash used, so any number here can be traced back to the exact code and data that produced it.