Interpretation of AI-Generated vs. Human-Made Images

Area of Science:

• Computer Vision
• Human-Computer Interaction
• Media Studies

Background:

• The proliferation of AI-generated content necessitates understanding its impact on audience perception.
• Distinguishing between AI-generated and human-made images is becoming increasingly challenging.
• Current research lacks comprehensive data on how image origin affects interpretation and evaluation.

Purpose of the Study:

• To investigate if the origin of an image (AI-generated vs. human-made) influences its interpretation.
• To explore the role of visual professionalization in the evaluation of AI-generated and human-made images.
• To assess audience perception of realism and credibility based on image source.

Main Methods:

• 161 participants (71 visual professionals, 90 non-professionals) evaluated 24 AI-generated and 8 human-made images.
• Image evaluation included source identification, perceived realism, and attributed credibility.
• Data collected across 5152 responses for each evaluation criterion.

Main Results:

• Human-made images were more accurately identified as such compared to AI-generated images.
• AI-generated images were frequently misclassified as human-made by participants.
• Human-made images received higher ratings for both realism and credibility than AI-generated images.

Conclusions:

• Individuals demonstrate limited accuracy in determining an image's origin (AI vs. human-made).
• Inaccurate source identification significantly affects the perceived credibility of visual content.
• Visual professionals and non-professionals alike struggle with source attribution, impacting image evaluation.