Comments on: Eliza, the Simulated Rogerian Psychotherapist. http://grandslacker.wordpress.com/2008/04/27/eliza/ The Archives Tue, 21 Apr 2009 00:54:35 +0000 http://wordpress.com/ hourly 1 By: slacker http://grandslacker.wordpress.com/2008/04/27/eliza/#comment-383 slacker Sat, 03 Jan 2009 06:08:38 +0000 http://grandslacker.wordpress.com/?p=154#comment-383 I chose Eliza because she/it is available in Javascript, which is supported by every modern browser. Eliza fooled the people she was tested on because they were not told to think that Eliza was a computer program. You expected Eliza to be a computer program and so your objectivity is tainted. Of course, to us, it is obvious that Eliza is a computer program because we are subjectively familiar with chatter-bots. The people who Eliza was tested on were subjectively unfamiliar with chatter-bots. This is another weakness in the Turing Test, the difficulty in defining the phrase "fooled a person". I chose Eliza because she/it is available in Javascript, which is supported by every modern browser.

Eliza fooled the people she was tested on because they were not told to think that Eliza was a computer program. You expected Eliza to be a computer program and so your objectivity is tainted.

Of course, to us, it is obvious that Eliza is a computer program because we are subjectively familiar with chatter-bots. The people who Eliza was tested on were subjectively unfamiliar with chatter-bots. This is another weakness in the Turing Test, the difficulty in defining the phrase “fooled a person”.

]]> By: focal http://grandslacker.wordpress.com/2008/04/27/eliza/#comment-382 focal Sat, 03 Jan 2009 05:34:46 +0000 http://grandslacker.wordpress.com/?p=154#comment-382 I find the program to be quite lousy. I could think myself some obvious simple rules to improve the speech. Consider this dialog: Eliza: (some question) Man: No Eliza: Are you saying no just to be negative? Man: No Eliza: Why you keep repeating yourself? At this point it becaomes obvious that this is a bot, and not a very bright one. I find the program to be quite lousy. I could think myself some obvious simple rules to improve the speech. Consider this dialog:
Eliza: (some question)
Man: No
Eliza: Are you saying no just to be negative?
Man: No
Eliza: Why you keep repeating yourself?
At this point it becaomes obvious that this is a bot, and not a very bright one.

]]> By: slacker http://grandslacker.wordpress.com/2008/04/27/eliza/#comment-327 slacker Tue, 14 Oct 2008 15:03:26 +0000 http://grandslacker.wordpress.com/?p=154#comment-327 Well, I would have to disagree because a Rogerian Psychotherapist is a real kind of person and thus, if such a simple program as Eliza can accurately impersonate such a Psychotherapist (and she did and can) then we have a serious plausibility challenge. Even if you are right, though, it does not discount any of the other problems with the Turing test, including the "Chinese Room" analogy. I'd be interested in hearing your objection to that, if any. Well, I would have to disagree because a Rogerian Psychotherapist is a real kind of person and thus, if such a simple program as Eliza can accurately impersonate such a Psychotherapist (and she did and can) then we have a serious plausibility challenge.

Even if you are right, though, it does not discount any of the other problems with the Turing test, including the “Chinese Room” analogy. I’d be interested in hearing your objection to that, if any.

]]> By: sokrates http://grandslacker.wordpress.com/2008/04/27/eliza/#comment-326 sokrates Tue, 14 Oct 2008 14:24:52 +0000 http://grandslacker.wordpress.com/?p=154#comment-326 The problem with Eliza is that the one that has to decide whether he/she is dealing with a real Rogerian therapist does not have a valid idea what the remarks and answers of a real therapist should be. Indeed the response of Eliza is a senseless repeating of the statements and remarks posed. You are bounced back to your own ideas, and you have to reflect on what you are saying. According to Rogers this is the best thing you can do. So the program is in this respect the best representation of a Rogerian therapist. However, it only proves that a Rogerian therapist can simply be simulated by a recursive computer program. The Turing test itself is not challenged at all. The problem with Eliza is that the one that has to decide whether he/she is dealing with a real Rogerian therapist does not have a valid idea what the remarks and answers of a real therapist should be. Indeed the response of Eliza is a senseless repeating of the statements and remarks posed. You are bounced back to your own ideas, and you have to reflect on what you are saying. According to Rogers this is the best thing you can do. So the program is in this respect the best representation of a Rogerian therapist. However, it only proves that a Rogerian therapist can simply be simulated by a recursive computer program. The Turing test itself is not challenged at all.

]]>