Twenty Questions

The original game, with human players, is played as follows:

One player chooses a person, place or thing (which should be known to all players). The others then, in turn, ask a question of the first; questions must be answerable with "yes" or "no". It's considered good form, however, to clarify a literal answer that would otherwise be misleading. For example, one of my friends recently asked me whether the person I was thinking of was in Rochester. She was actually in the town of Greece, our particular suburb, but I answered "Yes, she's in greater Rochester.

The great advantage of this game is that it can be played in almost any situation: In a car with one player driving, during a hike, sitting at home, via almost any online medium.

Sometimes the categories animal, vegetable, and mineral are used instead of person, place and thing. The guessers have 20 questions, and if they do not get the correct answer before they run out, the Answerer ('It') wins. Usually if a guesser figures out what 'It' was thinking of, they in turn become It.

This whole thing is more fun if you get unlimited questions and can use anything from any category (things such as 'light', 'numbers', and 'HAL' don't fit well into the traditional three categories). A very good geek game. When you get bored with 20 questions, move on to Questions.

Regarding the http://www.20q.net/ website mentioned above by Grocery Sack Boy, I can't quit thinking of the correlation between this idea and the idea of Everything2. The data base on 20Q accumulates information in order to make better guesses at the target in the user's mind. At the end of each game, the data base will ask you if you want to update the data base with your input. This is the attempt to hone the system in order to better ascertain the target in future games.

When you begin a new game and are sitting there trying to think of something to target in your mind (like trying to think of something to node), it will even offer you hints about things which need more work in the data base. You will see something like this at the start of a game:

Some objects that need exercise: a gypsy moth, a gas stove (for heating), a courier (messenger), a blast furnace.

Does this not remind you of what The Content Rescue Team attempts to do here?

It will give you hard data on where the data base currently is in this process of becoming a perfect player:

Current age of knowledgebase: 929879 games or 9 years, 3 months and 17 days. This version of the game has played 901474 games with a success rate of 34.92%. People are currently playing about 50 games per hour.

Your choices of targets are Animal, Vegetable, Mineral, Other, or Unknown. Once you get something in mind and start to play it will begin the questions. Your choices for the answer to each question are:

• Yes
• No
• Irrelevant
• Unknown
• Sometimes
• Maybe
• Probably
• Doubtful
• Usually
• Depends
• Rarely
• Partly

At first, as you answer, it will say "too many probably objects" until it feels as if it has enough information to start listing some possibilities. It will give you stats on how close it thinks each possibility is. If it feels fairly confident that it's got it, it'll take a guess earlier than the 20th question. If not, it will guess at the 20th question. If it misses, it will continue past 20 questions and guess whenever it feels confident again. You can chose to give up at any time after this.

Once the game is over, the data base will offer this choice:

If you believe that the information you have entered is correct, press UPDATE.

This is the crucial information it uses in its quest for perfection.

It will then give you what it feels are the most likely answers for those times you answered Unknown. It will also tell you what it learned that it did not previously know. If it disagrees with your answers, it will tell you what items it thinks are contradictions.

In a way, it seems a mirror image of this site. It is using human input in order to gain information so that it can better defeat other human users who play in the future. There really is no way to access the knowledge it has gathered other than play the game and see how well it is doing. Your favorite data base here, on the other hand, is using human input in order to gain information so that other humans in the future can readily find out things they did not know or, at least, be entertained or amused.

All work and no play make an AI a dull toy. But using input to more efficiently hose future users would be just the direct opposite of E2, wouldn't it? Hopefully.

Both data bases rely heavily on the veracity of the input offered. And there's the rub, eh?