Ben Garfinkel

It also seems to me as well, these classic sort of desirability arguments don’t bite very strongly. So again, the fact that these platforms really focus on discovering points of consensus seems to reduce risks around issues becoming politicized if the public gets really engaged in them. It also reduces this risk of tyrannical majorities. I think it also reduces this risk of extreme divergence from expert views. If it’s a point that you actually need to get, like pretty broad consensus around, I think there’s less reason to worry that, you know, something basically very, you know, wacky will come out of it.

I do still think, though, that it is sort of interesting to think about these classic arguments, you know, regardless of exactly the level of merit you think they have, and sort of look at, you know, vTaiwan and these other projects and sort of think about, you know, do these actually imply limits or do projects like vTaiwan actually show that there’s, you know, these serious flaws in these arguments which were harder to point out when the arguments abstract, but may be easier when you actually have concrete cases of things that work in spite of them.

And so I just have some questions for both Hélène and Audrey. So one is, you know, how seriously should we take these sorts of classic concerns about sort of downsides or limits to participation? So I know Hélène actually is, you know, her most recent book and her other work actually engages a lot with these questions, so I’m sure she has a lot of thoughts. Then, you know, more concretely, you know, what are the limits of initiatives like vTaiwan? Are there actually any domains where they might be ineffective or undesirable? You know, for example, would something like, you know, using vTaiwan to set monetary policy, would that actually be a bad idea, or is that actually, you know, something that could, you know, quite plausibly work?

And then just more generally, and sort of, you know, zooming out, what are the actual limits of participatory democracy? So regardless of, you know, what you think the limits are at the present day, with enough technological and institutional innovation, I’m curious especially whether Audrey thinks it might be possible to just almost completely move beyond representative democracy altogether, and sort of get the benefits of much higher participation without the, you know, some of the downsides or limits that people believe exist today. Wonderful. So, Audrey and Hélène, feel free to respond.

Yeah. I think there are also different issues where it’d be interesting to, you know, there’s like different potential pitfalls of it. So if I imagine, for example, something, you know, like vTaiwan being used in the United States, I imagine a bunch of failure modes for, let’s say, it being used to talk about reparations, (laughter) you know, being different than the potential failure modes for it being used to talk about something like the minimum wage, where in one case it’s, you know, an issue with huge value disagreements and huge disagreements about, you know, how you ought to interpret history and things like that.

And then something like the minimum wage, it’s, you know, this, you know, actually like quite technical issue where economists super disagree about, you know, like with a $15 minimum wage, you actually have employment effects. How large are the welfare impacts of employment effects? How large are welfare, you know? Like, and so there’s, yeah. And so there’s, you know, interesting things that you might have about how well can this actually integrate to technical academic disagreements. There’s concerns you might have about does it just break down if people have like very different values and it’s, you know, really polarized to begin with. And so I think different issues might be interesting test cases for different sorts of concerns.

Yeah. I’d also — it’s the not something which will be available in the next decade, but the long run sci-fi thing I suppose I’m excited about is if, you know, we ever get to the point where you can actually really communicate, you know, your values to machine learning systems. You can really have machine learning systems have a good understanding of what your preferences are. That seems like something that could really seriously overcome a lot of the issues with participatory democracy, like especially the time cost.

Like if you can in some way automate the process of giving feedback about, “Oh, I would like this, or I wouldn’t like this,” then that also seems like, you know, something we’re not about to have, but something in the long run we might look forward to that could really allow for much more, you know, meaningful representation of viewpoints in the political process.

Yeah. Thanks so much.

Same to you. (laughter)