Researchers equip Mario characters with social intelligence
Wed 3 Feb 2016
Scientists have designed artificial intelligence software to apply social skills [PDF], based on human thinking and behaviour, to popular video game characters such as Mario, Luigi, Yoshi and Toad from the Super Mario Bros franchise.
The German team, based at Tübingen University’s Cognitive Modeling group, explained that although smart systems are playing an increasing role in gaming, social intelligence is still lacking. ‘While artificial intelligence has made tremendous progress in the field of computer gaming, artificial social intelligence is frequently no more than a marginal feature in the shadow of theoretical applications,’ said one of the team’s developers Fabian Schrodt.
Led by Professor Martin Butz, the researchers taught the game characters to watch and communicate with each other, as well as to learn from their surroundings. This was achieved through attaching a cognitive control loop to each agent, integrating insights from cognitive science, linguistics and psychology.
Characters were programmed to develop a motivation system, inspired by the need for wealth, progress, and full health. These drives then triggered specific game events, for example with a stronger longing for wealth, Mario would strive to reach more coins.
A release this week detailed that Mario could ask Toad how he’d collected coins, learn from Toad’s advice and put it into practice. In team play, the system allowed the figures to work together to achieve common goals. In this instance, Mario and Toad could learn to team up and stand on each other’s heads to reach coins placed in high positions. The socially-adapted game also enabled characters to turn against one another, even if they originally had a friendly relationship, should they be insulted or intentionally hurt.
The scientists have released a video of the prototype game in which they equip Mario with artificial social intelligence. In the clip Mario is presented as a self-motivated character who appears to live in his environment. The gamer can interject and influence his actions by giving abstract orders or motivational instructions through speech control. Mario can even ask for help or further information from the gamer to achieve his mission.
The researchers hope to continue developing the system, particularly for application in education and in future integration with computer games, human-machine research and driving assistance.