Julian Togelius at the IT University of Copenhagen, Denmark, said that computer games have had an element of adaptability for decades.
"If you play well the game gets harder and if you are lousy it might get easier," New Scientist quoted him as saying.
Togelius and colleague Georgios Yannakakis want to take this adaptability one step further by creating games that "learn" to identify whether an individual is a fun-junkie or a challenge-seeker, and then tailor later sections to suit these tastes.
Two people might ultimately play very different versions of the game - but both should be satisfied by the experience.
For their research, the scientists altered the game Super Mario Bros, varying parameters such as the number and type of enemies and the size of gaps between platforms in response to how the players fared.
The game also records a player's moves, including how often they run and jump, and the time spent standing still.
Later, volunteers played two slightly different versions of the game and were quizzed about which version they found more challenging or predictable, fun or frustrating.
The researchers used algorithms to identify which particular suite of parameters is associated with different gaming experiences.
"If you die by falling too often down gaps that is indicative of frustration," said Yannakakis.
However, he said that the approach goes beyond "common sense" associations to uncover those that are not so readily apparent.
In Super Mario Bros, for example, hitting bricks to release coins or stomping turtle shells and throwing them - activities not necessary to accomplish the overall goal - positively correlate with a fun experience, says Togelius.