Are Prompts All You Need?: Chatting with <scp>ChatGPT</scp> on Disinformation Policy Understanding
Bibliographic record
Abstract
ABSTRACT ChatGPT has shown promise in assisting qualitative researchers with coding. Previous efforts have primarily focused on datasets derived from interviews and observations, leaving document analysis, another crucial data source, relatively unexplored. In this project, we address the rapidly emerging topic of disinformation regulatory policy as a pilot to investigate ChatGPT's potential for document analysis. We adapt our existing qualitative research framework, which identifies five key components of disinformation policy: context, actors, issue, instrument, and channel, to sketch out policy documents. We then designed a two‐stage experiment employing a multi‐layer workflow using a dataset with highly relevant policy documents from US federal government departments. Through iteratively developing and refining six different prompt strategies, we identified an effective few‐shot learning strategy that achieved 72.0% accuracy and a 70.8% F‐score with the optimal prompt. Our experimental process and outcomes explore the feasibility of using ChatGPT to support manual coding for policy documents and suggest a coding approach for conducting explicit document analysis through an interactive process between researchers and ChatGPT. Furthermore, our results initiate a wider debate on how to integrate human logic with ChatGPT logic, along with the evolving relationship between researchers and AI tools.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.035 | 0.204 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.002 | 0.003 |
| Scholarly communication | 0.003 | 0.008 |
| Open science | 0.003 | 0.006 |
| Research integrity | 0.002 | 0.003 |
| Insufficient payload (model declined to judge) | 0.010 | 0.004 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".