Abstract
LLMs are among the most advanced tools ever devised for understanding and generating natural language. Democratic deliberation and decision-making involve, at several distinct stages, the production and comprehension of language. So it is natural to ask whether our best linguistic tools might prove instrumental to one of our most important tasks involving language. Researchers and practitioners have recently asked whether LLMs can support democratic deliberation by leveraging abilities to summarise content, to aggregate opinion over summarised content, and to represent voters by predicting their preferences over unseen choices. In this paper, we assess whether using LLMs to perform these and related functions really advances the democratic values behind these experiments. We suggest that the record is mixed. In the presence of background inequality of power and resources, as well as deep moral and political disagreement, we should not use LLMs to automate non-instrumentally valuable components of the democratic process, nor be tempted to supplant fair and transparent decision-making procedures that are practically necessary to reconcile competing interests and values. However, while LLMs should be kept well clear of formal democratic decision-making processes, we think they can instead strengthen the informal public sphere—the arena that mediates between democratic governments and the polities that they serve, in which political communities seek information, form civic publics, and hold their leaders to account.
- “We should not use LLMs to automate non-instrumentally valuable components of the democratic process”
- LLM summarization has a mixed record of performance (propensity to hallucinate, poor job integrating widely distributed insights) which threaten to undermine values like equality and conciliation
- LLMs lack explicit, transparent, and reproducible decision making algorithms, which is a problem for the value of conciliation
- LLMs as representatives undermines the non-instrumental value of participation and the instrumental values of preference transformation and collective intelligence
- ” We should not be tempted to supplant fair and transparent decision-making procedures that are practically necessary to reconcile competing interests and values.”
- Formal democratic decision procedures must be fair, simple, secure, transparent, and contestable. LLMs cannot satisfy these criteria.
- LLMs are the wrong kind of tool for reconciling competing interests and values.
- “We think LLMs can instead strengthen the informal public sphere—the arena that mediates between democratic government and the polities that they serve, in which political communities, seek information, form civic publics, and hold their leaders to account.”
- LLMs can help provide the epistemic preconditions for democracies.
- Improve content moderation and reranking social media feeds to align with broader societal values
- Enabling citizens to navigate the digital public sphere in ways that discourage polarization and extremism
- Offer valuable information retrieval and simplification/translation capabilities for citizens who struggle with complex information
- Combat conspiracy theories with tailored counterarguments and facilitating more measured dialogue online.
- LLMs can help provide the epistemic preconditions for democracies.