Neura Market
llm-lobbyist logo

llm-lobbyist

Free

Large Language Models as Corporate Lobbyists

FreeFree tier
Type
Open Source
Visit Website GitHub (40 stars)

About llm-lobbyist

This open-source research code implements a proof-of-concept for using large language models (LLMs) to perform corporate lobbying activities. It leverages OpenAI's text-davinci-003 (and earlier text-davinci-002) to determine whether proposed U.S. Congressional bills are relevant to specific public companies, providing explanations and confidence levels. For relevant bills, the model drafts persuasive letters to bill sponsors advocating for changes to the proposed legislation. The repository includes a novel dataset of ground-truth relevance labels for benchmarking LLM performance on this task. The associated paper discusses both the potential positive and negative implications of AI-driven lobbying and the risk of AI influencing law without direct human intent.

Key Features

Determines relevance of U.S. Congressional bills to specific public companies using OpenAI's text-davinci-003
Provides explanations and confidence levels for relevance predictions
Drafts persuasive letters to bill sponsors for relevant legislation
Includes a new ground-truth dataset for benchmarking legislative relevance assessment
Outperforms simple baseline of predicting most common outcome (irrelevance)
Benchmarks older GPT-3 model (text-davinci-002) for comparison

Pros & Cons

Pros
  • Proof-of-concept demonstrates LLMs can perform complex policy analysis and persuasion tasks
  • Outperforms simple baseline on relevance determination
  • Open-source code and dataset enable reproducibility and further research
  • Addresses important ethical and societal questions about AI influence on law
Cons
  • Requires access to OpenAI API (paid) and thus not fully self-contained
  • Limited to U.S. Congressional bills and specific public companies
  • Experimental and not production-ready; results may vary
  • Older model (text-davinci-002) performed worse than baseline, indicating sensitivity to model capabilities

Best For

Corporate lobbying and policy influenceAutomated legislative monitoring for public companiesResearch on AI alignment and law as informationBenchmarking LLM performance on policy-related natural language tasksExploratory analysis of AI's role in augmenting human lobbyists

FAQ

What does LLM Lobbyist do?
It uses large language models to determine if proposed U.S. Congressional bills are relevant to specific public companies and drafts persuasive letters to bill sponsors for relevant bills.
Which LLM is used?
The code uses OpenAI's text-davinci-003 (and text-davinci-002 for comparison).
Is there a dataset included?
Yes, the repository includes a new ground-truth dataset (legislation_relevance_dataset_for_llm_evaluation_unbalanced.csv) for benchmarking relevance assessment.
Is this code production-ready?
No, it is a research proof-of-concept. The paper discusses potential positive and negative implications of taking the idea further.