Summary of Evidence Of a Log Scaling Law For Political Persuasion with Large Language Models, by Kobi Hackenburg et al.
Evidence of a log scaling law for political persuasion with large language models
by Kobi Hackenburg, Ben M. Tappin, Paul Röttger, Scott Hale, Jonathan Bright, Helen Margetts
First submitted to arxiv on: 20 Jun 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The paper explores the persuasiveness of large language models in generating political messages, raising concerns about their potential impact. To investigate this, researchers generated 720 persuasive messages on 10 U.S. political issues using 24 language models spanning several orders of magnitude in size. A large-scale survey experiment (N = 25,982) was conducted to estimate the persuasiveness of each model. The findings reveal a log scaling law, where larger models’ persuasiveness is characterized by sharply diminishing returns, and task completion (coherence and staying on topic) appears to account for their persuasive advantage. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large language models can now create political messages that are just as convincing as those written by humans, which has raised concerns about how much this persuasiveness will continue to increase with model size. To learn more, researchers created 720 persuasive messages on 10 U.S. political issues using 24 different language models, ranging in size from small to very large. They then tested these messages in a big online survey (with over 26,000 people) to see how well each model could persuade people. The results show that bigger models aren’t much better at convincing people than smaller ones. |
