An Artificial Intelligence Has Been Designed That Shows Its Decisions

Artificial intelligence technology is rapidly advancing towards becoming an indispensable part of our lives. Artificial intelligence such as Siri and Alexa are examples of artificial intelligence that many people can easily reach. Understanding the decision-making methods of artificial intelligence software and observing how they will react when they make mistakes is of great importance for the development of this technology.

PhD student Upol Ehsan from the School of Interactive Computing and his team developed an interesting project based on the idea that understanding how artificial intelligence works will increase trust in artificial intelligence.

The team chose the game called Frogger to use in the research because it was well suited for observing how an AI makes decisions. First of all, while playing the game, what decision and why they made before the moves were taken audibly. Then an artificial intelligence was trained using this data.

27e696f150710ba9ae001c079c45b55e77b4e0c7

Three different explanations were shown on the screen while playing the artificial intelligence game. One was written by a human, the other by artificial intelligence. The third was randomly selected from among human and artificial intelligence explanations. The people watching the game did not know who made which decision. Viewers were asked to look at the resolutions and choose which one looked more like a human decision. Audiences mostly preferred human decisions, but artificial intelligence decisions were chosen second.

In a second stage, the audience was asked whether they preferred short and simple decisions or long and detailed ones from the AI’s decisions. Detailed and longer explanations were more preferred.

Ehsan made the following statement about the research result: “This project is more about understanding human perceptions and observing artificial intelligence preferences than the development of artificial intelligence. Artificial intelligence, with all its abilities, should be able to reach everyone without discrimination. We need human-centered artificial intelligence.”

Related Posts

Leave a Reply

Your email address will not be published.