20Q

Starting if necessary extending the resources introduced in these slides, investigate the technique behind the 20Q game can it be (how) implemented in a computer game ?

20Q works by using question matrices and weighting analysis. The questions and the possible answers are all weighted in different ways allowing 20Q to guess closer each time you play as it takes into account previous games because it learns mainly by repetition. The website also uses the rough location, age, and gender of the player, giving it the added help of searching through other similar users and seeing what their thoughts were when they played.
20Q is an artificial neural network and as such is continually learning using techniques similar to humans and as such seems to reach conclusions as humans would by weighting the answers and using heuristics.
To a certain extent the technique behind 20Q could be used in games, an example would be applying it to enemies in a shooter. This could make the enemies more likely to use tactics making them harder to kill, however there is a certain limit as the player doesn’t want to reach a point where they can’t defeat the enemies because they are too clever. It may be better for the players overall enjoyment if they used the the 20Q neural technique but with restrictions. Restrictions which could be applied through difficulty settings allowing players to enjoy the game on their own level. 
Another problem would be the likelihood of the player playing the game after they have finished it, if there was a 2nd playthrough inbuilt in the game e.g. Borderlands, they are more likely to but most of the time they are not. In games where there are 2nd playthroughs the technique could add that extra element where you may have to adopt a totally new style of play.
Overall the technique could be applied and make game play a lot more exciting however there is a fine line between challenging but still fun and too difficult.

Table Reference
Robin Burgener, "Artificial neural network guessing method and game"