Would you let artificial intelligence (AI) mow your lawn? File your taxes? Defend you in a trial? 

To help people answer these types of questions, researchers at the Georgia Tech Research Institute (GTRI) have developed SatisfAI, a serious tabletop and digital card game that encourages players to consider how much agency they are willing to share with AI-based systems to solve real-world problems.

SatisfAI uses a player-judge mechanic to promote both introspection and consideration of the preferences of others, indexing real-world objectives against Max Neef’s model of fundamental human needs, and asks players to indicate their preferred level of autonomy, from choices such as “No AI,” “Assistant,” and “Surrogate.”

“We wanted to create a tool to better understand how people prefer to interact with AI,” said Moon Kim, a GTRI senior research associate and project co-lead. “We focused on two key questions: Would you use AI for a particular task? And if so, how much automation would you prefer? Our motivation was to elicit insights that could inform the design and alignment of AI systems to human needs.”

Originally supported by GTRI’s Independent Research and Development (IRAD) program between 2021 and 2023, the game has since been played by approximately 270 K-12 teachers and education professionals across Georgia. It has received enthusiastic feedback from educators, who value the meaningful conversations it sparks during and after gameplay.

One teacher shared the following: “What SatisfAI gives teachers is a humanizing experience of talking with each other about AI based on their values, a conversation that was likely often skipped over when ChatGPT was released upon the world.”

Another said: “SatisfAI causes you to think about what using AI really means and how it impacts you.” 

For classroom use, SatisfAI is currently being customized to help students explore the tradeoffs of AI, fostering critical thinking and encouraging intentional decision-making about their learner agency. Through gameplay and guided discussions, GTRI aims to equip teachers and students with valuable insights as part of preparing for an AI-driven future in education.

In each round of the game, which is playable in person and online, one player acts as the judge. The judge player is given a real-world objective and is asked to decide whether AI should solve this objective, and how much agency they prefer to retain.

For example, a scenario might state the objective of taking care of a sick family member, asking the judge to imagine what different solutions might exist for various levels of AI automation — from a robotic assistant, an intelligent medication dispenser, a medical information advisor, or possibly no AI at all — then decide which level they prefer. Other players then try to predict the judge’s choice, earning points when they match the judge, with the highest score winning.

“Ultimately, SatisfAI is not based on the assumption that more AI is naturally better,” said Ethan Trewhitt, a GTRI senior research engineer and project co-lead. “Instead, it investigates how much automation people want and where they want AI involved in their life.”Looking ahead, the researchers plan to expand the game’s reach to more K-12 partners across the state and beyond, with future considerations to extend the game to industry partners and government sponsors to further explore the tradeoffs and broader implications of AI adoption and its impact on human agency.

If you are interested in learning more about SatisfAI, please contact Moon Kim at moon.kim@gtri.gatech.edu or Ethan Trewhitt at ethan.trewhitt@gtri.gatech.edu.

Writer: Anna Akins 
Photos: Sean McNeil 
GTRI Communications
Georgia Tech Research Institute
Atlanta, Georgia USA

For more information, please contact gtri.media@gtri.gatech.edu. 

To learn more about GTRI, visit: Georgia Tech Research Institute | GTRI