Ben Garfinkel

好。嗯，非常謝謝。我看看能不能順利分享螢幕。好。我的評論會講得蠻短的，多留點時間給討論。不過我覺得有條線索值得稍微拉一下：在臺灣進行的許多數位民主計畫，像 vTaiwan，好像有個共同的主軸，就是要找機會超越，你知道，純粹的代議民主。也就是不再像唐鳳說的那樣，只是，你知道，每四年送出一點點資訊，而是真正積極讓公眾參與，讓他們在選舉之間更主動地形塑立法、形塑政策，而不只是投票。

但同時，也有一個非常悠久的思想傳統——其實我得說，Hélène 最近的書把它整理得非常好，某些方面也反駁了它。總之，有一個悠久的思想傳統在為限制公眾參與辯護，或者說在捍衛代議民主，對抗那些需要更積極參與的民主形式。所以我想，可以稍微探討一下，像 vTaiwan 這樣的直接民主倡議，究竟有沒有回應政治學者和政治理論家，你知道，三不五時提出的這些經典疑慮。

在深入之前，我想也值得先講一下最基本的論證，說明為什麼我們應該先強烈傾向認為公眾參與是好的。讓公眾更多參與政策制定和立法，最簡單的論證是：一般民眾對自己的偏好和關切，通常比他們的代表了解得多。所以我們應該很傾向於認為，如果讓公眾更多參與，讓他們的偏好和關切直接進入決策過程，那麼結果往往會是更能反映公眾偏好的政策；相較之下，你知道，只是送出幾個位元的資訊——你投給誰——然後你的代表，你知道，再設法去猜什麼對你好、你會想要什麼。

另外也有一些，你知道，可能相當有說服力的論證，說明為什麼更多參與對民主通常是好的。很多人直覺上覺得，人們參與民主本身就有某種內在價值；政策如果以某種更直接的方式，反映或來自受政策影響的人的行動或偏好，也是如此。如果公眾更積極參與政策的產出，決策也比較可能獲得公眾支持，或被視為具有正當性。參與的人愈多，往往就會產生群眾智慧的效果。讓大家盯著過程，或讓過程更公開透明，其實也能避免貪腐。讓人們真正在政府中扮演積極角色，也可能帶來正面的社會效果，能夠，你知道，影響像是團結感，或者，你知道，相互理解的感覺。

但與此同時，也有很多限制參與的經典論證。我把它們分成兩類。第一類是務實性的論證。它背後的想法基本上是：更多參與原則上是有價值的，但在，你知道，某些特定領域之外，實在很難做到。其中一個論證就是，大群體很難進行有意義的審議。如果你找，你知道，十個人坐下來，讓他們以尊重、真誠的方式分享彼此的觀點，並試著達成共識，直覺上這會比讓一百萬人做同樣的事，你知道，自然容易得多。

而且，比較容易請，你知道，比方說十個人、一百個人，花時間真正坐下來了解一個議題，想清楚各種反對意見和反駁，然後得出某種結論。你知道，某種意義上這就是設立代表的意義，你知道，這本來就是他們的全職工作；而要讓那些，你知道，有其他全職工作、興趣和煩惱的人花時間做這件事，往往很難。在某些情境下，如果人們，你知道，覺得自己個人的參與不會帶來多大差別，或因為各種原因覺得參與沒什麼吸引力，光是要讓他們有動機參與就很難。

另一方面，則是可欲性的論證。人們常提出的疑慮是，至少在某些領域，更多參與其實會導致更糟的結果。不只是參與很好但很難做到，而是你其實會希望，對人們在選舉之間參與的程度設下一定的限制。從，你知道，政治學者到，你知道，古代的政治哲學家，長久以來都有人認為，需要在不同程度上讓某些決策與公眾隔開。

當然有，你知道，像亞里斯多德那樣的經典著作，表達了對某種不受節制的民主的擔憂。當然還有像 James Madison 這些人，他們參與了，你知道，起草美國憲法，而且，你知道，非常刻意地，你知道，把他們在做的事定位為建立一個共和政體，而不是純粹的民主政體；這是基於一個想法：直接民主，或是非常大量的參與，如果不是——如果在選民意志和實際結果之間沒有某種程度的間接性，就會有某些風險。

人們提出的疑慮範圍很廣。一個經典的疑慮是，選民在某些議題上往往專業有限，或缺乏遠見；尤其在古代哲學家，或者，你知道，前幾個世紀的著作裡，常常有一種現在聽起來更讓人不舒服的觀念，就是菁英不知怎的就是，你知道，比較適合做，你知道，某些決定。這當然，你知道，就是柏拉圖這類人著作裡的那種脈絡。人們提出的另一個疑慮是某些議題被政治化的風險，也就是說，有些決定如果拿來進行大規模的公共辯論，就有可能變得極化，或以其他方式讓決策過程崩壞。

還有人們提出的關於多數暴政的經典疑慮。這個想法是，你自然會想保護自由和權利，而如果你太偏向民主，太偏向讓公眾在沒有各層間接性的情況下直接參與決策，自由被侵犯的機會就更高。這些疑慮很多已經隨著時間淡去。所以我想，你會發現很少有人還在為柏拉圖那類人對民主的批評辯護，而且我認為我們把其中很多看作是，你知道，建立在，你知道，一種強烈得沒有道理的菁英主義上，是正確的。

但同時，這些疑慮雖然不常被說出口，很多仍然反映在大多數民主國家政府的運作方式中。我想最明顯的例子大概就是司法獨立。在大多數民主國家，有一個非常強的原則：法官是被任命或選出，任期相當長。然後他們做的決定，選民其實沒辦法直接影響。以美國最高法院為例，最高法院大法官顯然是被任命，而不是由選民直接選出，而且是終身職。所以他們基本上，你知道，一坐就是一輩子，基本上不受任何民主問責的影響。

這類制度設計，你知道，有時會受到質疑，但往往被視為理所當然；我覺得這顯示很多人至少隱約有這樣的觀念：你知道，有些領域最好和公眾保持一點距離。比方說，沒有多少人會積極主張讓公眾直接投票決定什麼合憲、什麼不合憲，或直接投票決定最高法院案件的結果。

這一脈也有一些比較新的，你知道，論證，帶有比較實證的性質。我想最極端的版本是去年出的一本書，叫《10% Less Democracy》，作者是經濟學家 Garett Jones，他基本上主張，那種麥迪遜式的，你知道，試圖限制民主或限制參與的想法，確實有它的道理。也有人舉出一些例子，作為支持這個立場的可能證據。

另一個非常經典的例子是，大多數經濟學家傾向認為央行獨立是好的。民選官員不能很直接地影響貨幣政策是好的，央行官員不由選民直接選出也是好的。這裡的思路是，如果讓民主對央行有更多控制，就比較可能導致印鈔這類短視的決定，而這又比較可能導致通膨失控。這方面也確實有一些實證證據，顯示獨立的央行似乎真的比較能管控通膨——至少如果我們假設通膨是你不想要太多的東西。

看起來，任命的法官在某些指標上表現也比民選法官好，包括公正性的指標。例如在州的層級，任命的法官比較少顯示出偏袒本州居民或本州選民、而不利於外州人的判決傾向。政治人物在接近改選時，也似乎比剛連任或即將退休時，更可能投下違背學界共識的票。比方說，政治人物比較可能反對像租金管制這類政策——這類政策有時很受歡迎，但大多數經濟學家認為會有嚴重的負面間接效果——當然，像租金管制這類政策究竟是不是真的不好，也還可以爭論。

另一個，你知道，非常顯著的例子是，最近在 Donald Trump 的彈劾審理中，有幾位共和黨參議員投票贊成彈劾，而非常引人注目的是，這些人包括幾位即將退休的參議員，幾位剛連任、所以六年內不必面對改選的參議員，還有兩位選區非常穩固的參議員。也有人暗示說，你知道，根據他們對當時發生的事相當正確的認知，以及，你知道，他們自己的良知，本來會有更多人投票支持彈劾；但由於某種，像是，缺乏——對，我想是缺乏——與選舉政治的隔離，他們比較傾向往一個方向投票，而那個方向至少從主觀角度看，似乎是不明智的。

也有一些證據顯示，至少在美國，當媒體開始報導某些政治議題時，可能會加劇極化，反而讓國會更難就此通過基於共識的立法。所以至少有這些，我想，是比較近期提出的論證。我覺得去思考這些論證對數位民主意味著什麼，還蠻有意思的。

在我看來，剛才討論到的一些倡議，像 vTaiwan 和公共政策網路參與平臺（Join），似乎其實部分反駁了一些反對參與的務實性論證。你知道，一個非常基本的點就是：這些倡議確實存在。它們確實，你知道，讓臺灣大量的人參與，而且似乎真的產出了良好、合理的結果，例如以一種大致能讓廣大群體接受的方式，來規範 Uber 在臺灣的立法。而且看起來，vTaiwan 倡議所用的一些工具，確實解決了擴大審議規模的一些經典問題。

舉些例子，我想前面已經提過了：這個倡議用軟體，根據贊成和反對的投票模式，自動描繪出同意與不同意的所在。它試著找出人們聚集在哪裡，或同意與不同意的各個不同面向，這讓處理大量一開始沒有結構的意見變得非常有效率，然後讓大家真正掌握整體樣貌，真正看到，你知道，我們在哪裡同意、在哪裡不同意。它取消了直接回覆別人的功能，減少了某種酸民問題。

它基本上會把有廣泛共識的留言或問題往上推。如果一個問題或建議好不好，看起來沒有強烈的極化，它就會被更顯著地凸顯出來，這降低了那種，你知道，大家一關注就更極化的風險。整件事似乎讓過程變得有效率得多。

在我看來，那些經典的可欲性論證在這裡也沒什麼殺傷力。再說一次，這些平臺真正專注於發掘共識點，這似乎降低了公眾一旦高度投入、議題就被政治化的風險。它也降低了多數暴政的風險。我想它也降低了與專家意見極端分歧的風險。如果某個點你真的需要取得，像是，相當廣泛的共識，我想就比較沒有理由擔心，你知道，會冒出什麼基本上非常，你知道，離譜的東西。

不過我還是覺得，思考這些經典論證挺有意思的——你知道，不管你認為它們到底有多少道理——然後去看，你知道，vTaiwan 和這些其他計畫，想一想，你知道，這些論證是否真的意味著某些限制；還是說像 vTaiwan 這樣的計畫其實顯示了，你知道，這些論證有嚴重缺陷——這些缺陷在論證很抽象的時候比較難指出，但當你真的有具體案例，看到某些東西儘管如此仍然行得通時，也許就比較容易指出來。

所以我有幾個問題想請教 Hélène 和唐鳳。第一個是，你知道，我們應該多認真看待這些關於參與的缺點或限制的經典疑慮？我知道 Hélène 其實在，你知道，她最近的書和其他著作裡，大量處理這些問題，所以我相信她有很多想法。然後，你知道，更具體地說，你知道，像 vTaiwan 這樣的倡議有什麼限制？是否真的有某些領域，它們可能無效或不適合？你知道，例如，像是，你知道，用 vTaiwan 來制定貨幣政策，那真的會是個壞主意嗎，還是其實，你知道，是某種，你知道，相當有可能行得通的事？

然後更廣泛地，某種程度上，你知道，拉遠來看，參與式民主的實際極限在哪裡？不管，你知道，你認為現今的限制是什麼，只要有足夠的技術和制度創新，我特別好奇唐鳳是否認為，也許有可能幾乎完全超越代議民主，某種程度上得到更高度參與的好處，而沒有，你知道，人們認為今天存在的一些缺點或限制。太好了。那麼，唐鳳和 Hélène，兩位請隨意回應。

對。我覺得不同的議題也有不同的有趣之處，就是有不同的潛在陷阱。比如說，如果我想像 vTaiwan 這樣的東西在美國被使用，我能想到的失敗模式，用它來討論賠償問題（笑）和用它來討論最低工資這類議題，是很不一樣的。一種情況是價值觀有巨大分歧，對於如何詮釋歷史等等也有巨大分歧。

而像最低工資，你知道，其實是一個相當技術性的議題，經濟學家之間意見分歧很大：比如說，15 美元的最低工資確實會產生就業效應。就業效應的福利影響有多大？福利影響有多大？等等。所以有一些有趣的面向，像是這種做法到底能不能有效整合技術性的學術爭議。還有人會擔心，如果人們的價值觀非常不同、一開始就已經極化了，這種做法會不會就此崩潰。所以我覺得不同的議題可能是不同類型疑慮的有趣測試案例。

對。我也會——這不是在未來十年內就會有的東西，但長遠來看，那個比較科幻的方向，我想我對此感到興奮的是，如果，你知道的，我們真的到了那一天，你可以真正地，你知道的，把你的價值觀傳達給機器學習系統。你可以真正讓機器學習系統充分理解你的偏好。那似乎真的能大幅克服參與式民主的許多問題，尤其是時間成本。

就像，如果你能以某種方式把提供回饋的過程自動化——「噢，我想要這樣，或者我不想要那樣」——那似乎也是一種，你知道的，我們不會馬上擁有、但長遠來看可以期待的東西；它真的能讓更多觀點在政治過程中得到有實質意義的代表。

對。非常感謝。

你也是。（笑）

Great. Yeah. Thank you so much. Let me see if I can successfully share my screen. Great. So I will keep my comments pretty short, just to leave more room for discussion. But a thread I thought it’d be interesting to pull on a little bit is that it seems like, you know, sort of a unifying theme of a lot of the digital democracy projects happening in Taiwan, like vTaiwan, is that they’re aiming to look for opportunities to move beyond, you know, pure representative democracy. So no longer, as Audrey put it, just, you know, sending a bit of information every four years, but really actively engaging the public, and giving them a more active role in shaping legislation and shaping policy between election years, not just voting.

There’s also at the same time a really long intellectual tradition, which actually, I should say Hélène’s recent book does a really good job of summarizing, in some ways refuting. But there’s a long intellectual tradition defending limits to public participation or defending the idea of representative democracy against forms of democracy that involve more active participation. So I thought it’d be interesting to explore a little bit how some direct democracy initiatives like vTaiwan address or don’t address these sorts of classic concerns that political scientists and political theorists have, you know, brought up from time to time.

Before diving into that, I think it’s also just worth making the sort of, you know, very basic case for why we should have a strong prior that public participation is good. And really simplest argument for involving the public more in policymaking and legislation is that members of the public typically know a lot more about their own preferences and concerns than their representatives do. So we should have a strong inclination to think that if you involve the public more, and their preferences and concerns feed directly into decision-making processes, then we should often expect the outcomes to be policies that better reflect the public’s preferences relative to, you know, just sending some bits of information, you know, who did you vote for, and then your representatives, you know, try and figure out what would be good for you or what you would want.

There are also some other, you know, potentially pretty compelling arguments for why more participation tends to be good in democracy. So a lot of people have the intuition that there’s something intrinsically valuable about people participating in democracy and policies actually in some more direct way reflecting or flowing from the actions or the preferences of the people who are affected by the policies. Decisions are more likely to achieve public buy-in or be perceived as legitimate if the public participates more actively in producing policies. If you involve more people, you often get wisdom of the crowd effects. You can actually avoid corruption by having people keep an eye on the process or have the process be more open and public. There’s also potentially positive social effects of having people really actively have a role in their government, that can, you know, affect things like a sense of unity or, you know, a sense of understanding.

At the same time, though, there are also lots of classic arguments for limits on participation. I’ll divide these into two categories. One are pragmatic arguments. The idea behind these is basically that greater participation would in principle be valuable, but it’s just really hard to achieve outside of, you know, certain specific domains. So one argument is just it’s hard for large groups to meaningfully deliberate. If you take, you know, a group of ten people and you sit them down, you have them sort of share each other’s viewpoints in a respectful, good faith way, and try to come to consensus, there’s some intuition this is a naturally, you know, easier thing to do than having a million people do this.

It’s also easier to ask, you know, let’s say ten people, a hundred people to take the time to really sit down and learn about an issue and think through the objections and counter-objections and come to some sort of conclusion about it. You know, this is sort of in some sense the point of representatives is, you know, it’s their whole job to do this and it’s hard often to get, you know, people who have, you know, other full-time jobs and interests and concerns to take the time to do this. In some contexts it can also be hard just to get people motivated to participate if they, you know, don’t feel like their individual participation will make that large of a difference or it’s unappealing for various reasons.

Then on the other hand, there are desirability arguments. So there are also concerns that people often put forward that at least in certain domains, greater participation would actually lead to worse outcomes. It’s not just that participation is desirable but hard to achieve, but that you actually want to have certain limits on how much people participate between elections. There’s a really long history of people, from, you know, political scientists to, you know, ancient political philosophers, perceiving that there’s a need to insulate certain decisions from the public to differing degrees.

So there’s of course, you know, classic writing by people like Aristotle expressing concern about sort of unmitigated democracy. Of course people like James Madison who are involved in, you know, drafting the Constitution of the United States, you know, were very pointedly, you know, trying to frame what they were doing as building a republican government as opposed to a purely democratic government, based on the idea that there were certain risks with direct democracy or really heavy involvement that’s not — that doesn’t involve some degree of indirectness between the will of voters and what actually happens.

And there’s a range of concerns that people have brought forward. So a classic concern is the idea that voters often have limited expertise on certain issues or lack farsightedness, especially in the context of ancient philosophers or, you know, work from previous centuries is often a much more uncomfortable notion that elites are just somehow, you know, better suited to making, you know, certain decisions. This of course, you know, this sort of strain in the work of people like Plato. Another concern people bring up is a risk of politicizing certain issues in some sense that there’s certain decisions where, if you have a large public debate about them, there’s some risk that it’ll become polarized or in some other way decision-making process will break down.

And there’s classic concerns people brought up about tyrannical majorities. The idea that you naturally want to protect liberties and rights and if you lean too far into democracy, if you lean too far into giving the public direct involvement in decisions without various levels of indirectness, then there’s a higher chance of liberties being violated. A lot of these concerns have faded over time. So I think you’re gonna find very few people defending the sorts of critiques of democracy people like Plato had, and I think we rightly view a lot of them as, you know, based on, you know, an unjustifiably strong form of elitism.

At the same time though, a lot of these concerns, even if not voiced very frequently, are still reflected in the way that the governments of most democracies work. So I think the most obvious example of this is probably judicial independence. In most democracies, there’s a really strong principle that judges are appointed or elected for like quite long terms. And then the decisions they make, you can’t really be directly influenced by voters. So in the context of something like the US Supreme Court, obviously Supreme Court justices are appointed as opposed to directly chosen by voters and they have lifetime appointments. So they’re just basically, you know, fixed for life and they’re basically insulated from any sort of democratic accountability.

And these sorts of institutional designs are, you know, are sometimes questioned, but they’re often taken for granted in a way that I think suggests a lot of people at least implicitly have the notion that, you know, there’s some areas where it’s good to have a bit of insulation. There aren’t that many people who actively defend, for example, the idea of having a public directly vote on what’s constitutional or not constitutional or directly vote on the outcomes of Supreme Court cases.

There’s also some more recent, you know, arguments in this vein have had a more empirical character. So I think the most extreme version of this is a book that came out last year called 10% Less Democracy by an economist, Garett Jones, who basically argues that the sort of Madisonian, you know, idea of trying to place limits on democracy or limits on participation does actually have some merit to it. And there’s some examples that are brought up as potential evidence in favor of this position.

So another really classic case is that most economists tend to think it’s good for central banks to be independent. It’s good for elected officials to not be able to very directly influence monetary policy and it’s good for central bank officials to not be directly elected by voters. And yet the thought process here is that if you allow for more democratic control over central banks, then this is more likely to lead to short-sighted decisions about printing money, which are more likely to lead to runaway inflation. And there is some empirical evidence here that central banks that are independent do actually seem to manage inflation better, at least if we assume that inflation is a thing that you don’t want to have that much of.

It seems to be the case that appointed judges perform better on certain metrics than elected judges, including metrics of impartiality. For example, at the state level, judges who are appointed show less evidence of bias in terms of deciding in favor of people who are residents of the state or voters in the state versus out-of-state people. Politicians also seem to be more likely to vote against academic consensus when they’re close to reelection as opposed to when they’ve just been reelected or they’re about to retire. So politicians are more likely to, for example, oppose policies like rent control, which are sometimes popular but most economists think would have serious negative indirect effects, although it can of course be debated whether, you know, policies like rent control actually in fact are bad.

And so another just, you know, very salient example of this is recently during the impeachment hearing for Donald Trump, a number of Republican senators voted to impeach, and it’s really striking that this was made up of a couple of senators who are about to retire, then a few senators who have just been reelected and so won’t have to face reelection for six years, and two senators who are in very safe seats. And there’s some suggestion that, you know, more people would have voted for impeachment, but based on plenty of accurate beliefs about what happened in that context and, you know, their own consciences, but due to sort of, like, lack of, yeah, I guess, insulation from electoral politics, they were more inclined to vote in a direction, at least from a subjective standpoint, seems ill-advised.

There’s also some evidence as well that when media starts to cover certain political issues, at least in the United States, it can increase polarization, and it can make it actually harder for Congress to pass consensus-based legislation on it. So there’s at least the, I think the arguments which have been more recently made. And I think it’s sort of interesting to reflect on what these arguments imply for digital democracy.

It seems to my mind that some initiatives like vTaiwan and Join, which have just been discussed, seem to actually partly refute some of the pragmatic arguments against participation. So you know, one just, you know, very basic point is that these initiatives exist. They have in fact, you know, involved huge numbers of people in Taiwan, and they have actually seemed to produce outcomes which are actually good and sensible, like legislation regulating Uber in Taiwan in a way that’s like broadly amenable to large groups of people. And it does in fact seem like some of the tools that the vTaiwan initiative uses actually do resolve some of the classic issues with scaling up deliberation.

So just as some examples, which I think have already been mentioned, the initiative uses software to automatically map points of agreement and disagreement based on upvoting and downvoting patterns. It tries to figure out where people cluster or sort of the different dimensions of agreement and disagreement, which makes it really efficient to take large amounts of sort of initially unstructured inputs and then give people actually a lay of the land and actually see, you know, where do we agree and disagree. It removes the ability to reply directly to people, which reduces sort of trolling issues.

It essentially upvotes comments or questions, which there’s a broad consensus around. If there doesn’t seem to be a strong polarization around whether a question or suggestion is good, it gets highlighted more strongly, which reduces this risk of, you know, if people pay attention, they’ll get more polarized. And just the whole thing seems to make the process much more efficient.