Re:人也不是理性的动物

所有跟贴·加跟贴·新语丝读书论坛http://www.xys.org/cgi-bin/mainpage.pl

送交者: mangolasi 于 2006-4-27, 10:51:27:

朝三暮四的例子举得很不好，因为它完全可以通过新古典的思路去解释：我们当下的享受和迟下的享受带来的效用是不同的。“今朝有酒今朝醉”，带来的满足感是比留待明天才饮的要大，因此明天的享受是要打“折扣”的。我们通常是用dynamic programming来表达，两餐总共带来的享受是：u(c_1)+\beta*u(c_2)，那\beta就是少于１的“折扣”了。如果你的c_1+c_2是七个生果，而u()是一个单调递增函数，你会分配得c_2比c_1大么？至于是４－３还是７－０，这是由你的u()有多大“非线性”来确定。

而那个不赌的理智，更是为了“非线性”来做注脚，恰恰这“非线性”正是我们理性的体现（新古典的看法）。无饭食的时候，一碗白饭也是珍贵，而对着山珍海味，多一碟龙虾也不觉得多兴奋－－这个“边际”并不是一成不变的，我们生来的感受（未涉及到决定，于是与理智无关），倒是作决定时照顾到我们这个特质反而是理性的。“边际”随着手中的钱变而变，于是多一元和少一元，对我们的效用是不同的。恰恰一个正常人不仅是非线性，还是concave的，少一元的影响，比多一元的影响要大，于是我们是讨厌风险的－－并非只是由于恐惧。你要我玩那个游戏，必须要赢的钱比输的钱要多，如果概率相等的话。当然，2.5和1的比较有点悬殊，讨厌风险的理性特质未必可以全部解释，而且，他们试验是用两组人来比较，因此思路是对的，不过试验结果要我来解释，我会说是讨厌风险的程度的不同的原因，也说得通－－于是只是改一改concavity的不同，理性人的假设仍然可以继续用。

但“理性的做法应该是每次都赌，而不应该弃权，因为每次输赢的概率相等，而赢的回报却高于输的损失，对玩家有利”很有问题，理性不是每次遇到风险就算算期望回报（我们应该看回报带来的满足，而不是回报per se），拿最大那个。只有"财雄势大”才会这样做－－我们于是可以假设机构的u()是线性。“杰出的投资者之所以能大获成功”，我也可以说是古典的原因：杰出的投资者的u（）是线性的，或者是天生如此，或者是因为有钱：有钱的人，多一元和少一元带来的满足程度的改变相差不大。

行为经济学有一个很大的问题：试验室里的试验都是“小儿科”，stake不大，数据能不能extropolate都很成问题。有一些人不做试验，反而是观察电视上的game show的数据，倒是一个可能的出路，因为game show的stake，倒是比较真。

所有跟贴:

人也不是理性的动物, 方是人, and so he 不是理性的. - email (0 bytes) 2006-4-29, 08:42:21
理性也有不同的层次，就像控制rational decision making的方程有不同的order一样 - conner (64 bytes) 2006-4-27, 16:21:18
So much for today. I need to go home right now. - mangolasi (0 bytes) 2006-4-27, 15:51:14
Enlighten, Mr.Fang's gambling example is still flawed even for 20 times. - mangolasi (1020 bytes) 2006-4-27, 14:47:50
- 方说的这个实验目的非常明确 - eng (554 bytes) 2006-4-27, 15:21:19
  - 不要折腾了，心理学实验有很多时候是不能向研究对象说明研究目的 - xj (34 bytes) 2006-4-27, 17:49:58
    - 研究目的和游戏目的是两回事 - eng (113 bytes) 2006-4-27, 21:38:53
  - 由我从方先生的帖子看出，这个试验非常有问题，除非它的方法更subtle。 - mangolasi (1114 bytes) 2006-4-27, 15:45:39
    - 人家测的是心理，$20还达不到要分析财富的程度 - xj (119 bytes) 2006-4-27, 17:53:15
    - 我想方的意思是 - eng (232 bytes) 2006-4-27, 15:54:52
      - 那么方先生的理性定义和绝大部分人都很不一样。 - mangolasi (119 bytes) 2006-4-27, 16:00:52
        
        这个明显是不同的风险 - eng (159 bytes) 2006-4-27, 16:16:41
        
        我有很强的信心这些参与者是在酒足饭饱的情况下去做实验的:) - eng (0 bytes) 2006-4-27, 16:09:52
        
        Gee, cantonese, you leaked more cantonese in your posting more - steven (22 bytes) 2006-4-27, 16:04:26
- I just realized that too. Maybe Fang's example should be - Enlighten (262 bytes) 2006-4-27, 14:52:01
  - Just like Americans say "you are opening a whole can of worms" - Enlighten (0 bytes) 2006-4-27, 14:55:37
别给朝三暮四这个故事添新意了 - Enlighten (489 bytes) 2006-4-27, 14:41:48
- 添新意也不是不可以 - eng (138 bytes) 2006-4-27, 14:46:05
  - 具体点行吗？方用这个例子表明猴子不懂加法交换律，这有什么争议吗？ - Enlighten (0 bytes) 2006-4-27, 14:48:09
    - 你连我是支持你还是反对你都看不出来？ - eng (28 bytes) 2006-4-27, 14:59:02
      - Not your fault. It's just that I am so used to the American style of - Enlighten (103 bytes) 2006-4-27, 15:50:48
你说的例子恰好证明了人既是理性的也是非理性的 - eng (440 bytes) 2006-4-27, 14:40:56
- 凡推理做决定都有一个“信”的开头。 - mangolasi (993 bytes) 2006-4-27, 15:23:48
  - 能够用数学模型模拟的并不一定是理性的 - eng (489 bytes) 2006-4-27, 15:42:45
    - 那是proposition。而理性只是涉及到有了proposition后的决定过程。 - mangolasi (214 bytes) 2006-4-27, 15:50:34
      - 真实不是理性 - eng (316 bytes) 2006-4-27, 16:07:28
only HunHunSheng got the point - 008 (935 bytes) 2006-4-27, 14:20:47
- "Monkey's got no banks and they aren't care about delays" - mangolasi (338 bytes) 2006-4-27, 14:33:17
  - Lack of computational power - 008 (304 bytes) 2006-4-27, 14:45:13
    - I am not Steven Wolfram. - mangolasi (336 bytes) 2006-4-27, 15:56:44
- You don't know what you are talking about. - Enlighten (35 bytes) 2006-4-27, 14:32:49
  - Why should I read? It is about - 008 (28 bytes) 2006-4-27, 14:39:16
    - It is about "emotional" instead of "rational" - 008 (0 bytes) 2006-4-27, 14:46:01
Fang is correct in using 朝三暮四的例子 - Enlighten (207 bytes) 2006-4-27, 13:52:49
- No. Time discount factor to our utility means "impatience", - mangolasi (779 bytes) 2006-4-27, 14:09:18
  - The story might as well have been the other way around: - Enlighten (349 bytes) 2006-4-27, 14:29:06
    - Monkey said: i like 3+4 better than 4+3. - 008 (165 bytes) 2006-4-27, 14:35:46
    - I have nothing to say if the story is told in the other version. - mangolasi (0 bytes) 2006-4-27, 14:34:42
  - should we start from assumption or from the fact - 008 (276 bytes) 2006-4-27, 14:26:41
不错，但是这样一来，就不是“通俗”了。 - mirror (0 bytes) 2006-4-27, 11:18:40
- 镜先生，不是每个人都有你那么道行高深的。 - mangolasi (48 bytes) 2006-4-27, 11:24:26
  - 就拿“朝三暮四”来讲，经济学的效用说可以“破”， - mirror (116 bytes) 2006-4-27, 11:49:45
    - Maybe it is better for monkeys to - HunHunSheng (161 bytes) 2006-4-27, 12:14:22
      - Monkeys eat 4 accorns in the morning, because they 早上要练长跑。 - Latino2 (0 bytes) 2006-4-27, 13:00:40
        
        Wait. Does monkey eat acorns? - mangolasi (75 bytes) 2006-4-27, 13:10:53
        
        Are you still running? - HunHunSheng (101 bytes) 2006-4-27, 13:09:28
        
        Any reason for why a DaZi eats 4 tomatos in the morning? - 笑笑 (0 bytes) 2006-4-27, 13:07:27
        
        如果前15%里面有个女金丝猴，朝5暮2也一样。 - Latino2 (0 bytes) 2006-4-27, 13:05:27
      - Are you saying the evolution story is totally wrong? - gyro (0 bytes) 2006-4-27, 12:16:07
        
        no. he''s just like the guy who claimed that the - Enlighten (44 bytes) 2006-4-27, 12:20:41
your 非线性 is wrong too. here should use "average" - Enlighten (138 bytes) 2006-4-27, 11:05:45
- yup, 20 times is a decent number, this is a point I missed. - mangolasi (478 bytes) 2006-4-27, 11:12:10
  - It would be too much to ask for a short article that targets - Enlighten (19 bytes) 2006-4-27, 11:16:02
    - Slight modification can avoid giving out such an impression. - mangolasi (218 bytes) 2006-4-27, 11:22:01
      - It takes a book to explain utility - gyro (141 bytes) 2006-4-27, 11:29:59
        
        Then you need a few line in a simpler book. - mangolasi (167 bytes) 2006-4-27, 11:34:07
        
        Fang's utility function is good persuasive effect on the general public - gyro (98 bytes) 2006-4-27, 11:40:46
why dynamic programming? a simple time discount is enough. - Enlighten (0 bytes) 2006-4-27, 11:00:23
- This is the most basic dynamic programming. - mangolasi (58 bytes) 2006-4-27, 11:04:12
  - but all we need here is to prove 3+4 is worse than 4+3. - Enlighten (49 bytes) 2006-4-27, 11:08:28
    - "prove 3+4 is worse than 4+3" simple time discount is fine, - mangolasi (124 bytes) 2006-4-27, 11:14:41
      - If you can explain something by using a simpler - Enlighten (58 bytes) 2006-4-27, 11:22:52
        
        isn't dynamic programing a simple well-known approach? - steven (0 bytes) 2006-4-27, 11:56:55
        
        I can assure you that this programming terminology is misleading and stupid - xj (63 bytes) 2006-4-27, 12:25:19
        
        it was way before computer time, it was after linear programming. - steven (0 bytes) 2006-4-27, 12:59:31
        
        Blame the rocket scientists (literally). - mangolasi (0 bytes) 2006-4-27, 12:43:27
        
        Richard Bellman was an mathematician. - steven (0 bytes) 2006-4-27, 13:01:35
        
        An applied one. And I have the impression that DP is developed - mangolasi (57 bytes) 2006-4-27, 13:06:21
        
        And the word "stochastic control" is still prevailing for SDP. - mangolasi (0 bytes) 2006-4-27, 13:07:12
        
        I mean, the most important part of SDP. - mangolasi (0 bytes) 2006-4-27, 13:07:42
        
        whatever it is, 99% people don't know anything about linear programming - xj (26 bytes) 2006-4-27, 13:10:00
        
        The concept is very simple. - mangolasi (68 bytes) 2006-4-27, 13:11:58
        
        普通人基本上不关心概念，但我同意方没有写好这一篇 - xj (756 bytes) 2006-4-27, 13:35:29
        
        Dissent. - mangolasi (2145 bytes) 2006-4-27, 14:29:06
        
        we two are talking about different things - xj (683 bytes) 2006-4-28, 00:15:45
        
        You are wrong on two things. - Enlighten (360 bytes) 2006-4-27, 13:50:58
        
        你中文写作不过关。好的故事就是旧典故翻新意 - xj (912 bytes) 2006-4-27, 14:24:36
        
        not for general public. time discount is much better known. - Enlighten (63 bytes) 2006-4-27, 12:10:38
        
        Philosophically, time discount in utility and time discount in interest - mangolasi (545 bytes) 2006-4-27, 12:21:19
        
        Making things simple is the most difficult thing to do. - mangolasi (0 bytes) 2006-4-27, 11:26:38

加跟贴

所有跟贴·加跟贴·新语丝读书论坛http://www.xys.org/cgi-bin/mainpage.pl