TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Eigenmorality

439 点作者 bdr将近 11 年前

21 条评论

derefr将近 11 年前
As Aaronson points out, PageRank has a few edge-cases when used to do this analysis, basically because it treats its graph as a closed, internally-solipsistic system--it has no definition of morality other than what each of its nodes prefer of one-another. This works if you have a diverse spectrum of preference functions distributed among the nodes (the result tends toward a &quot;live and let live&quot; meta-ethics), but if your analysis is aimed at a preferentially homogeneous group (e.g. Nazi Germany), PageRank won&#x27;t give you the solution of &quot;move the &#x27;evil&#x27; majority toward the tenets of the good minority.&quot; It&#x27;ll instead suggest that the optimal system would have the &#x27;good&#x27; minority give up and become &#x27;evil&#x27;.<p>Scott Alexander suggests (<a href="http://slatestarcodex.com/2014/06/20/ground-morality-in-party-politics/" rel="nofollow">http:&#x2F;&#x2F;slatestarcodex.com&#x2F;2014&#x2F;06&#x2F;20&#x2F;ground-morality-in-part...</a>) you could instead use DW-nominate, the tool that does meta-cluster-analysis to mathematically detect &quot;party lines&quot; in congress (which are basically just clusters in human-utility-function-space anyway), to find what preference-subfunctions (e.g. helping old ladies cross the street, returning a wallet you find laying on the ground) correlate together into a cluster (that might be called &#x27;goodness&#x27;) -- and then grounding&#x2F;normalizing the PageRank analysis with that, such that you can tell whether the system as a whole is in a &#x27;good&#x27; or &#x27;evil&#x27; state.
评论 #7926201 未加载
knowtheory将近 11 年前
This is <i>very long</i> but worth reading.<p>The modeling exercise herein is basically attempting to use a game theoretic model to test out some really dumb&#x2F;simplified models of cooperation and whether the behaviors observed approximate anything approaching what our intuitions might say is moral behavior, up to and including an &#x27;eigenjesus&#x27; and &#x27;eigenmoses&#x27; up against tit_for_tat bots and the like.
评论 #7925779 未加载
评论 #7925982 未加载
jonnathanson将近 11 年前
Please note that what follows can be interpreted as criticism, but it&#x27;s not intended as such. I found this article quite interesting, and for me, it was the starting point for a lot of different thoughts about game-theoretical approximations of morality. So what follows is a somewhat tangential addition to the article, and not a critique of it.<p>My problem is not with the &quot;eigenmorality&quot; concept, nor with the various takes on playing it out across consecutive Prisoner&#x27;s Dilemma sessions. That aspect is extremely interesting. Rather, my problem is with the Prisoner&#x27;s Dilemma as a valid ground on which to test something like morality.<p>The Prisoner&#x27;s Dilemma is a foundational, theoretical framework for evaluating human behavior. And it&#x27;s a wonderful, elegant framework. But it treats humans as emotionless agents, and the &quot;punishment&quot; as an abstract, theoretical, rationally navigable scenario. Place real human beings into the Prisoner&#x27;s Dilemma, with real-world consequences, and you get all sorts of unexpected results. The Prisoner&#x27;s Dilemma is notorious for holding up perfectly fine <i>in vitro</i>, but less so <i>in situ</i>. Cultural conditioning plays a <i>huge</i> role in how real people act in the game. So do emotions, and irrational heuristics like overemphasizing loss aversion. (Tversky and Kahneman&#x27;s work has a lot to say about the latter.)<p>Using the Prisoner&#x27;s Dilemma as a proving ground, I think you&#x27;d arrive at an abstract model of morality -- but you wouldn&#x27;t capture how morality actually plays out with quasi-rational, emotional, circumstantially driven, human agents. And, philosophically speaking, that&#x27;s where morality actually counts the most.
评论 #7926890 未加载
评论 #7926223 未加载
MichaelDickens将近 11 年前
This is an interesting idea. Aaronson may be joking when he says he&#x27;s &quot;solv[ed] a 2400-year-old open problem in philosophy,&quot; but in case he&#x27;s not, this doesn&#x27;t come anywhere close to solving ethics. Philosophically speaking, it&#x27;s still necessary to show why his definition of &quot;moral&quot; holds up. All he&#x27;s done is assess a certain quality and then call it &quot;morality.&quot; I think it could better be called &quot;meta-cooperativeness&quot; or something like that.<p>I think Aaronson realizes this, because he does talk about how Eigenjesus and Eigenmoses don&#x27;t accord with our moral intuitions in some cases. He also addresses this somewhat in the section &quot;Scooped by Plato.&quot; His major point--that something like Eigenjesus can be useful, even if it cannot deduce terminal values--still holds.
评论 #7927769 未加载
andrewflnr将近 11 年前
I think the definition of morality in the article is far too simplistic. In my (Christian) view, it&#x27;s an important aspect of moral maturity to be able to be <i>nice</i> to immoral people without <i>cooperating</i> with their goals. Besides that dichotomy, the article already mentions that the model lacks critical information, specifically, the actors don&#x27;t know whether the other actors they&#x27;re [not ]cooperating with are &quot;good&quot; or &quot;bad&quot;.<p>That said, I find this approach to defining morality fascinating. Maybe if the definitions are refined it will manage to tell us something we already know (not entirely sarcastic; that would be legitimately impressive for a mathematical construct regarding morality).
评论 #7926043 未加载
MarkPNeyer将近 11 年前
my friends and i had started on this already. i had a hard time explaing to people why it was valuable; looks like scott has done it for us.<p>please help us!<p><a href="https://github.com/neyer/dewDrop" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;neyer&#x2F;dewDrop</a><p>right now all we have is a way to state which facebook users a person trusts. there&#x27;s a chrome extension to help with this. it&#x27;s extremely basic.<p>i have a server running at <a href="https://dewdrop.neyer.me" rel="nofollow">https:&#x2F;&#x2F;dewdrop.neyer.me</a> - we need a lot more help!<p>i&#x27;m just putting it on github now - so i&#x27;ll update the readme in a few minutes.
评论 #7925875 未加载
评论 #7926224 未加载
评论 #7926047 未加载
minority将近 11 年前
Considering that a majority of people who agree with each other are &quot;moral&quot; is highly problematic. Even if everyone in the system is morally equal, this system would automatically create and enhance differences between groups.<p>The author uses the example of climate-change deniers to express the opinion that minority groups have &quot;withdrawn itself from the main conversation and retreated into a different discourse.&quot;<p>Is this true of other minority groups - feminists? Homosexuals? Minority ethnic groups? It seems highly awkward to claim the same thing.<p>A better system would be one which considers how to cater for individuals rather than declaring a populist majority to be a special, protected ingroup. There&#x27;s enough of the latter already.
评论 #7927811 未加载
MichaelDickens将近 11 年前
This seems related to the idea of coherent extrapolated volition (<a href="https://intelligence.org/files/CEV.pdf" rel="nofollow">https:&#x2F;&#x2F;intelligence.org&#x2F;files&#x2F;CEV.pdf</a>). Both have some of the same problems--in particular, setting up the system requires making moral judgments about how to do so, so it&#x27;s not actually value-neutral.<p>(Aside: If I have two completely different thoughts about an article, should I post them in two separate comments or in the same comment?)
评论 #7926417 未加载
mrb将近 11 年前
Wow: <i>&quot;The mathematical verdict of both eigenmoses and eigenjesus is unequivocal: the 98% are almost perfectly good, while the 2% are almost perfectly evil.&quot;</i> The author says this diverge violently from most people’s moral intuitions, but actually this result is PRECISELY what moral relativism predicts. See, there are 2 school of thoughts attempting to explain where morality comes from:<p>- either morality is an absolute concept (things are inherently good or evil, theists might say this good&#x2F;evil is defined by a god or gods). This is <a href="http://en.wikipedia.org/wiki/Moral_absolutism" rel="nofollow">http:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Moral_absolutism</a><p>- or morality is relative, defined by people, defined by cultures (what one culture might consider immoral, another culture will consider it moral, and nobody is inherently right or wrong). This is <a href="http://en.wikipedia.org/wiki/Moral_relativism" rel="nofollow">http:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Moral_relativism</a><p>If moral relativism is right, it would be absolutely expected that the 98% are &quot;almost perfectly good&quot;, since they do things that the majority consider good. What a fantastic essay...
评论 #7926206 未加载
评论 #7926271 未加载
评论 #7927755 未加载
jzwinck将近 11 年前
&quot;The deniers and their think-tanks would be exposed to the sun; they’d lose their thin cover of legitimacy.&quot;<p>Don&#x27;t we have the ability to do this now by visualizing or analyzing citations? A set of &quot;fake&quot; think-tanks which promote bogus ideas should be identifiable as a mostly-disconnected component of a graph today. We don&#x27;t need to get each think tank&#x27;s explicit opinions about the others. Aaronson points out this single-purpose inquiry would encourage gaming, but analyzing a graph built for other incentives may give more &quot;honest&quot; results (at least for a while).<p>And we have, at least five years ago: <a href="http://arstechnica.com/science/2009/01/using-pagerank-to-assess-scientific-importance/" rel="nofollow">http:&#x2F;&#x2F;arstechnica.com&#x2F;science&#x2F;2009&#x2F;01&#x2F;using-pagerank-to-ass...</a> . You can follow links from there to a project called EigenFactor, academic research about shortcomings of PageRank in this application, and more.<p>Results of such analyses should be used as input to human thought processes and not some sort of legislative robot.
ipsin将近 11 年前
I found the addendum about the time-sequence of bad acts to be the most interesting, in that how you approach the problem leads to another wide spray of outcomes.<p>Scott mentions the &quot;forget the past&quot; and &quot;address root causes&quot; sides, but how do you deal with things in the middle?<p>Even being able to provide a model that allows for injustices from centuries ago would be impressive, but how should such things decay? Again, the same pressures come into play, based on the interests of the judged parties.
MakeUsersWant将近 11 年前
It&#x27;s probably no coincidence that repeated prisoner&#x27;s dilemna models another phenomenon: willpower.<p>George Ainslie argues in &quot;Breakdown of Will&quot; that will is actually the result of negotiations between past and future selves.<p><a href="http://www.picoeconomics.org/HTarticles/Bkdn_Precis/Breakdown_Will.pdf" rel="nofollow">http:&#x2F;&#x2F;www.picoeconomics.org&#x2F;HTarticles&#x2F;Bkdn_Precis&#x2F;Breakdow...</a>
hhm将近 11 年前
This Tolkien quote builds a similar circular definition of &quot;worth&quot;, which might be amenable to the same kind of analysis. <a href="https://twitter.com/JRRTolkien/status/480127254857400320" rel="nofollow">https:&#x2F;&#x2F;twitter.com&#x2F;JRRTolkien&#x2F;status&#x2F;480127254857400320</a>
评论 #7925783 未加载
bryan_rasmussen将近 11 年前
It seems to me that this would only be of interest if it can be shown that an immoral person is not someone that cooperates with other immoral people but not with moral people.
评论 #7925876 未加载
cma将近 11 年前
Needy babies are moral monsters according to many of these models..
评论 #7926210 未加载
tveita将近 11 年前
His definition is much closer to &quot;popularity&quot; than to anything I would recognize as &quot;morality&quot;.<p>It&#x27;s strange to exclude intent from your model when it&#x27;s an important factor in almost all systems of morality.
neotoy将近 11 年前
Good read, but I can&#x27;t help but end in thinking that by the time all of this would have been figured out, our civilization will be long gone.
yason将近 11 年前
There is no right or wrong, just acts with unescapable consequences and your freedom to learn something from your choices.
评论 #7926489 未加载
hyperion2010将近 11 年前
You can&#x27;t &quot;solve&quot; this problem in the same sense that you cannot develop a universally consistent foundation for mathematics. Goedel is there preventing you from EVER proving that one set of axioms is better than another.<p>I again wrote a longer response but have shortened it because the author seems to have committed a rather grave error which is to assume that human moral &#x27;intuition&#x27; is in any way consistent. There are heaps of evidence (cue the trolley car) that human moral judgements really should not be considered a guide for anything. The fact that we can capture the disasters of collective morality observed under various regime&#x27;s during the 20th century ought to tell us that following those models as a universal foundation for human relations is a terrible idea.<p>Might also be worth paying a visit to eigennicolo and not adhere to such rigid systems.
评论 #7927997 未加载
lohankin将近 11 年前
I was following Scott&#x27;s posts for a while. Most notable feature of those posts: everything he says is predictable. Blog is designed to appeal to the liberal academic establishment, which knows answers to all important questions, and is never in doubt. I don&#x27;t remember a single example of Scott&#x27;s opinion which could be deemed controversial in any sense. &quot;Eigenconformism&quot; would be a better name for his blog.
评论 #7926476 未加载
javert将近 11 年前
Happiness is the only intrinsic value for a human being, and thus a moral person is a person who pursues happiness effectively. (How to do that is another story.) However, Aaronson&#x27;s proposed definition of a moral person is not the effective way to pursue happiness. Thus, it is immoral.<p>It&#x27;s also immoral to call for all of us to sacrifice industrial output for future generations to solve the supposed climate change problem. There is no reason to presume that future generations are more important than the present generation (in fact, it is demonstrably the case that they are not). Thus, this position is profoundly immoral.<p>However, the implicit assumption that sacrifice is moral is common to most world religions and also altruism, which is probably where he imported it from. All of them are morally bankrupt. A scientist shold be able to be skeptical and see such logical flaws, even if he is not able to propose the correct solution.
评论 #7927338 未加载
评论 #7927336 未加载