Literature Review
Day 2 - Block 3
(Literature) search is hard
… especially if there are paywalls
Prompt: "Give me a literature table on greenwashing in top-tier marketing journals, with methods and key findings."
It might still work though (kind of)
| Journal | Paper | Method | Key finding |
|---|---|---|---|
| Journal of Consumer Research | Newman, Gorlin, Dhar (2014) | Experiments | Green initiatives backfire when consumers infer self-serving motives; perceived hypocrisy drives negative evaluations. |
| Journal of Marketing | Luchs et al. (2010) | Experiments | “Sustainability liability”: green products are seen as less effective in strength-related categories → indirect driver of skepticism. |
| Journal of Consumer Psychology | Forehand & Grier (2003) | Experiments | Firm motive attribution → skepticism; ulterior motives reduce persuasion (core mechanism behind greenwashing perceptions). |
Keyword != Semantic search
Some tips
Prompt for notification about paywalls
Otherwise fails silently and reports based on abstract
Provide list of journals
Quality ranking of research is weak so far (GPT likes researchgate, because its accessible!)
Activate web-search
SOA models do this automatically; However, monitor whether it actually searches the web. Reduces hallucinations!
False results
We are aware of false positives (“hallucinations”)
(i.e. papers that do not exist but are “found”)
For literature search, false negative are a huge problem too!
(i.e. papers that do exist but are not “found”)
Impose more structure on search
Research Rabbit
Build a literature agent: Tools
arXiv MCP:
- Let AI search and download arXiv papers
Zotero MCP:
- Automatic RAG on papers!
- Citation via bibtex-key
⚠️ Copyright
Don't
- Closed-source papers to LLM
- Trust AI citations blindly
- Closed-source papers to LLM
- Trust AI citations blindly
Do
- Log Paywall blocks
- Use private AI models
- Log Paywall blocks
- Use private AI models
References
TI Workshop | D2 B3