LLMs' Potential Influences on Our Democracy: Challenges and Opportunities

A growing literature [Potter et al. 2024, Rozado 2024, Feng et al. 2023, Santurkar et al. 2023, Hartmann et al. 2023, Vijay et al. 2024] shows that LLMs exhibit left political leaning on a range of issues. This has sparked active discussion about LLMs' potential influences on political discourse and democratic processes. Recent papers have taken initial steps forward in exploring this question. For example, Potter et al. (2024) showed that interactions with LLMs can sway voters towards the Democratic candidate, even without LLMs being instructed to persuade voters. Fisher et al. (2024) revealed that LLMs with radical political leanings can inject their political views into users' decisions on less sensitive political topics. However, these studies only examined the immediate effects of LLMs, leaving the question of their long-term effects open. On the other hand, Costello et al. (2024) demonstrated how LLM conversations can be used positively to durably reduce users' conspiracy beliefs

LLMs' Potential Influences on Our Democracy: Challenges and Opportunities --> --> LLMs' Potential Influences on Our Democracy: Challenges and Opportunities Yujin Potter 1 , Yejin Choi 2 , David Rand 3 , Dawn Song 1 1 UC Berkeley 2 University of Washington 3 MIT --> * Corresponding author --> * Several authors have unlisted affiliations in addition to their listed university affiliation. This piece solely reflects the authors' personal views and not those of any affiliated organizations, including the listed universities. --> Summary A growing literature [ Potter et al. 2024 , Rozado 2024 , Fen

related reading

• [2306.13000] Apolitical Intelligence? Auditing Delphi's responses on controversial political issues in the USarxiv.org
• [2303.17548] Whose Opinions Do Language Models Reflect?arxiv.org
• Discovering Language Model Behaviors with Model-Written Evaluations — LessWronglesswrong.com
• Andrew B. Hallandrewbenjaminhall.com
• How A.I. could help democracy.slate.com
• Assessing Political Bias in Language Models | Stanford HAIhai.stanford.edu
• Non-Determinism of “Deterministic” LLM Settingsarxiv.org
• Whoever Controls Language Models Controls Politics – HANNES BAJOHRhannesbajohr.de
• Against LLM Reductionism — LessWronglesswrong.com
• The persona selection model — LessWronglesswrong.com
• Probe-Based Data Attribution: Surfacing and Mitigating Undesirable Behaviors in LLM Post-Traininggoodfire.ai
• Six ways that AI could change politics | MIT Technology Reviewtechnologyreview.com