Vox Civitas: AI-Powered Civic Action
Bibliographic record
Abstract
Although artificial intelligence has advanced rapidly, there are still low participation levels, biases, and inability to scale up with civic participation and policy discussion, most digital participation platforms have less than 35 percent response rates, and fewer incorporate citizen views into practical policy results. This paper is a research proposal of an AI-based civic action system, incorporating large language models (LLMs), natural language processing (NLP), and computational social choice algorithms to enable civic deliberation that is large and inclusive, and data-driven. The framework was experimented with 3,000 simulated citizens profile based on the knowledge of participatory design, AI governance, and algorithmic fairness by applying sentiment analysis, topic clustering, and evaluation of the quality of the argument to mediate online discussion. In all stages of simulation and alignment, we used the GPT-3.5-Turbo model as our base LLM. The 3,000 synthetic citizen profiles were generated by conditioning GPT-3.5-Turbo on demographic, policy-interest, and sentiment constraints, and cross-validated with accuracy of 90% using a publicly available participatory-budgeting dataset (OpenPB-2022) to guarantee plausible distribution." The experimental findings show that the accuracy of consensus detection was improved by 27%, the moderate time was diminished by 41%, and the F1-score was 0.89 to detect a common-ground statement, as well as 22% of fairness and inclusivity when using the tool compared to the standard deliberation tools. Our improvements over a RoBERTa-large baseline for moderation and a TF-IDF consensus voter were statistically significant (paired t-test, p<0.05) across 5 runs. These results affirm that it is true that AI-human facilitation can be effectively scaled up to deliberative democracy, without impairing the level of representation or transparency. Expanding on it, the paper presents a new concept, the CivicGPT, the dynamic deliberation engine, that combines the work of the LLM and participatory budgeting with fairness optimization, which is a radical change in the system of ethical and transparent and participatory decisions made by the population.
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 | 0.004 | 0.008 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.001 | 0.003 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.007 | 0.002 |
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".