AI-enabled deliberative democracy

The proponents of “deliberative democracy” have spent decades dredging a this-but-that argumentative quagmire that has yielded nothing of either theoretical or practical value for democracy. One of the prolific underlying springs of sticky material for the quagmire has been the inherent contradiction between two dicta of “deliberative democracy”: mass participation and deliberation. It is very straightforward that masses cannot deliberate. Meaningful deliberation can occur in groups of at most a few hundreds of people (and even at this scale all-to-all deliberation could occur only under very favorable conditions).

Thus, “deliberative democracy” professionals can develop entire careers stirring, pouring and piling the sands of participation and deliberation without ever managing (or, it could be argued, without ever trying) to build any solid structure. Those of us who would suggest that both mass participation and deliberation are at best tools for good outcomes, rather than sacrosanct goals, are severely chastised for looking for illegitimate “shortcuts”.

Technology is one of the implements that have been routinely used to stir those sticky sands. Over and over again we have been promised that new information technology would allow democracy to go where it has never gone before. Mass education, remote participation, virtual mass discussions, crowd-sourced documents – these and many other unprecedented tools of democracy would be enabled by innovative technology. The fact that such promises go back to the advent of the radio (and probably much farther back) never discourages the prophets of mass participation from promising that the next technological innovation would be the one that would usher in democratic utopia where millions of voices would be heard by millions of people who all make meaningful – equally meaningful – contributions to decision making.

In the spirit of the times (or maybe a bit behind the times), the latest technology recruited to the cause of mass deliberation is Artificial Intelligence. The advantage of AI over previous technologies is that it is essentially magic. Unlike radio, TV, phones, video conferencing, or any other specific device, AI has no well-defined function. It can, we are to assume, do whatever humans do but without human limitations or bias. (And can do it all on the cheap!) AI can be an impartial moderator and facilitator of group discussions, an instant translator, an accurate fact-checker, it can cluster and organize arguments, track all the exchanges anyone has had with anyone else and measure the degree of overlap in content, track the arch of the conversations, analyze the similarities in content and counting up the number of new ideas, measure the quality of deliberation (stepping up its facilitator or fact-checking functions when quality drops), take cognitive pictures of the group, share consensus across groups, and seed assemblies with high-potential ideas.

(It almost seems like with all this magic, human-like, or super-human functionality, we should just sit back and have the AI tell us what to do. But then that would be an illegitimate shortcut, depriving the citizens of the sacred privilege of mass deliberation.)

By the way, for those who are a bit concerned that AI would be taking on too much power, in case something goes awry, e.g., if AI decides that anti-immigration sentiments and anti-minority views are “high-potential” ideas, an ethics committee would be there to set things right.

What could go wrong with a black-box hi-tech system that picks and chooses between speakers, decides which claims are true and which are false and which ideas are “high-potential” and should be emphasized, decides when a deliberation should be stopped because it produces no more “new ideas”, or that it should take a new direction because it is a “low quality” deliberation?

The image at the top of the post was generated by the “Nightcafe” AI text-to-image generator from the text “deliberative democracy”.

One Response

  1. […] France. Academics have continued publishing papers and opinions on the pros and cons of sortition (unfortunately often rehashing very well hashed material) but applications of sortition have been fading in […]


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: