As artificial intelligence becomes more sophisticated, prevalent, and capable of producing more realistic images and text, it has become critical for the public to distinguish \u201creal\u201d texts and images from \u201cnon-real\u201d ones.<\/p>\n
That is the premise behind two apps created by Lindsay Grace, Knight Chair in Interactive Media and professor at the University of Miami School of Communication, who has devoted decades to designing video games and teaching media literacy.<\/p>\n
He was recently awarded the first-ever lifetime achievement award from IndieCade, an\u00a0international juried festival of independent games that celebrates creativity, innovation, and artistry in all forms of interactive media.<\/p>\n
The two apps, Factitious and FactImage, are digital games that challenge the players by presenting them with news stories and photographs, which they must identify as either created by humans or by artificial intelligence. The apps recently won the Gee! award for best learning games. The award recognizes excellence in educational game design and is judged by international leaders in the field.<\/p>\n
\u201cI have been working on the problem of real-to-fake media\u00a0for three years, and I wanted to create a suite of apps to help discern misinformation from images and texts,\u201d he said.<\/p>\n
Grace added that fake information can be extremely dangerous for individuals and society by undermining trust, damaging reputations, and deepening social and political divisions.<\/p>\n
The apps, which can be downloaded from the Apple App Store and Google Play for Android devices, are available in several languages, including English, Spanish, Portuguese, Korean, and Chinese.<\/p>\n
Both apps reward the player with a number of points, according to the accuracy of the answer and the speed with which the question is answered. If a player reaches a score of 80 or higher more images or stories will be loaded, up to a maximum of more than 400 pictures, said Grace.<\/p>\n
FactImage offers players 30-second rounds. Each round presents a series of image cards. Players can reveal the image source for a clue\u2014at a point cost\u2014while learning the same reading and provenance-checking techniques used by professional fact-checkers. Grace keeps data on player performance for future research.<\/p>\n
After each card, a detailed reveal explains whether the image is created by a human or AI.<\/p>\n
Some of the clues that reveal whether an image is generated by AI are:<\/p>\n
distorted hands
\nimpossible backgrounds
\ninconsistent lightning
\nblurry faces in the background
\nincongruent setting<\/p>\n
\u201cEventually\u2014as AI improves\u2014these will be corrected,\u201d said Grace. \u201cBut your tendency to look for these clues will help you identify them faster.\u201d<\/p>\n
One of the images shown in FactImage shows a scene of a dog charging a cat who is climbing up a light pole. A man is walking in the background.<\/p>\n
The image may fool an uneducated eye, but Grace said that there are many clues that give it away as a fake image, including the almost perfect texture of the dog\u2019s fur, the blurring bottom of the light pole, and the dog\u2019s demeanor, which does not reflect an aggressive or curious attitude.<\/p>\n
For Factitious, players are provided stories from various news sources. Grace provides articles from credible sources such as The Guardian or BBC News, along with AI-generated content.<\/p>\n
\u201cWhen you are trying to detect AI-generated texts, you have to read between the lines,\u201d said Grace.<\/p>\n
Grace said that some clues to detect fake stories are:<\/p>\n
Unaccredited sources
\nEm dashes
\nUnnecessary quotations
\nWrongly worded names of institutions or groups
\nRun-on sentences
\nSentences with too many commas, used as connectors<\/p>\n
Factitious provides nearly 500 articles, evenly split between fake and real.\u00a0Currently, the accuracy rate across all players is 45 percent, said Grace.\u00a0<\/p>\n
To further increase the reach of media literacy, Grace has published a book called \u201cThe Literate Citizen: A Guide to Communication Literacy in a Networked World.\u201d<\/p>\n
He said it embodies his beliefs that \u201cin an age of deepfakes, algorithmic content streams, and fractured information ecosystems, traditional media literacy no longer meets the demands of our communication environment.<\/p>\n
Analyzing a news article or deconstructing an ad is only part of the skill set needed today. We are not merely media consumers but active participants in a vast, evolving exchange of messages.\u201d<\/p>\n
Grace is disseminating news about the apps through social media and at conferences. He also hopes that educators and others can learn more about media literacy. To that end, he has developed a guide for teachers.<\/p>\n