# Resources

Ordered by trust. The browsable version is at `reference/resources.html`.

## Primary source

- **Thinking, Fast and Slow**, Daniel Kahneman (2011). Part I for the System 1 and System 2 framework, then the heuristics and biases chapters in Part II. Check section references against your copy.

## Peer-reviewed research

- Goddard et al., systematic review of automation bias (PMC7651899). Lessons 0001, 0003, 0010.
- Sharma et al., sycophancy in language models (arXiv 2310.13548, ICLR 2024). Lesson 0007.

## Cognitive surrender research

- Shaw and Nave, Thinking, Fast, Slow, and Artificial (SSRN 6097646, 2025). Working paper, Wharton. Lessons 0001, 0010.
- Interactive explainer of Shaw and Nave by Avinash Mohan Chetty (datachutney.io). Secondary.

## Preprints (verify before citing)

- Huang et al., anchoring effect across LLMs (arXiv 2505.15392). Lesson 0002.
- Valencia-Clavijo, mechanistic anchoring study (arXiv 2511.05766). Lesson 0002.
- When Wording Steers the Evaluation (arXiv 2601.13537). Lesson 0005.
- More or Less Wrong, MathComp benchmark (arXiv 2506.03923). Lesson 0005.
- Provenance density and the transparency penalty (arXiv 2609.03460). Lesson 0006.
- Metacognitive decoupling and AI explanations (arXiv 2603.29681). Lesson 0008.
- To Rely or Not to Rely (arXiv 2412.15584). Lesson 0008.

## Secondary (lower trust)

- LessWrong wiki, Thinking, Fast and Slow. Orientation only.
- Wikipedia: Automation bias; Representativeness heuristic. Check citations.
- Ulrich Boser, 5 Tips To Keep Your Brain From Surrendering To AI (Forbes, 2026). Lesson 0010.
- A 2025 Science study on sycophantic AI responses (2,405 participants). Reported second-hand; verify before citing.

## Communities

- Peers inside Improving who evaluate AI output in client work.
