Ideological Turing Test: Playtest report

by Joshua_Fox1 min read7th Mar 2018No comments


Ideological Turing Tests
Personal Blog

We enjoyed an Ideological Ideological Turing Test game at LessWrong Tel Aviv and then again at LessWrong Jerusalem in February.

ITT has been played very rarely before. The Columbus Rationality group once did it. Their approach was to vote on the actual stance of authors of essays written by non-group members, gathered before the meeting.

We refined the game as we played it in Tel Aviv. If you play it, please let me know how you refined the rules.

Our rules:

  • A list of topics is written on the board. People volunteer to go up and speak.
  • They can choose from the board or find their own issues.
  • They speak 90 seconds pro, 90 seconds con.
  • In each case they take 20 seconds to prep and to avoid tipping their hand as to which they prefer.
  • Count votes from the audience:
    • Was the speaker convincing for pro? for con?
    • What is the speaker's actual stance?
  • The author then reveals their actual stance.

Example topics:

  • Veganism
  • Drugs legalization
  • Han Unification
  • Post-Modernism
  • Homeschooling
  • Scientology
  • AI Foom


1. Gamification:

To get a true game, you might give points for those who fool the audience about their actual stance; or you could award points, ice-skating style, for good form.

2. Convincing vs role-playing

Are speakers trying to be convincing, or to accurately simulate the position of someone who believes in the position?

Most LWers, with their rationalist bent, prefer the former. For this, you will need well-reasoned arguments.

The latter is the original ITT, as described by Bryan Caplan. I prefer this version: I want to learn the Dark Arts.

For example, if you were arguing pro post-modernism, to be convincing you would explain, like Scott Alexander, that agendas shape understanding of truth. To simulate a true believer, you would mix in some babble. But sound sincere. Poe’s Law aside, if you go overboard, you will fool no one!

3. Knowing the speaker’s opinions

A topic where the audience knows the speaker’s position, say, Scientology, makes a fun challenge for the speaker, but it becomes impossible to gamify the guess about the speaker's position.

On the other hand, if you do not know the speaker’s position, you can have fun guessing, but such topics are usually less controversial and so less interesting.


New Comment