Artificial Intelligence-Generated Images Challenge Human Perception

Perception researchers at Johns Hopkins University have developed novel artificial intelligence-generated images that appear to be one thing but reveal a different form when rotated. These "visual anagrams" hold promise for studying human visual perception and have already yielded insights into how people process real-world size effects.

Key Takeaways:

• The team created visual anagrams using an AI tool, which include images that are both a bear and a butterfly, an elephant and a rabbit, and a duck and a horse when rotated.
• The researchers found evidence for classic real-world size effects, such as preferring images of objects to be depicted in their ideal size, using visual anagrams that are rotated versions of the same image.
• The team used visual anagrams to study how people process size effects, finding that subjects adjusted the bear image to be its ideal size, making it bigger than when they adjusted the butterfly image, even though the two images are the same.
• The researchers believe that visual anagrams can be used to study animate and inanimate objects, and expect the technique to have many possible uses for future experiments in psychology and neuroscience.
• The team's findings suggest that visual anagrams can provide a more uniform stimulus for studying visual perception, addressing a longstanding need for consistent testing stimuli.

Statistics:

• The researchers created a total of 3 visual anagrams, including a bear and a butterfly, an elephant and a rabbit, and a duck and a horse.
• The team's experiments found that subjects adjusted the bear image to be its ideal size, making it 23% larger than the adjusted butterfly image.
• The researchers expect visual anagrams to be used in future studies to investigate a range of visual perception effects, including animacy, emotion, and shape perception.

Sources:

• Current Biology
• Johns Hopkins University
• Perception & Mind Lab, Johns Hopkins University