Our research study on Scrabblesque has gone live. Please take a few moments to play the game. You may even find that it is fun to play!
It turns out that even “simple” word games like Scrabble have a lot of inherent complexity. In the context of artificial intelligence, creating computer opponents that are believable is a fundamentally different problem than creating computer opponents that are optimal. In many ways, it is much more difficulty, given that humans are not particularly rational in the first place.
The game also allows us to cognitively model player actions. For example, there are a lot of interesting questions in this domain, of which I list only a small subset:
- When and why do people shuffle before selecting their final word?
- Do people change their word selection before submitting?
- When do people swap tiles?
- How often do people play optimal words? hat are time intervals between word selection?
- Can a human player identify whether their opponent is a human player or a computer?
Even data concerning user interfaces in games is not easily available. For instance, we may wish to know how often a player makes an error due to a button being to small or being placed in an awkward location. Scrabblesque allows us to perform this kind of HCI analysis.
So please help out and put some words on the board!
Discussions
