AI Hallucination and Distorted Information Classification

Overview of AI Hallucination

• AI hallucination refers to instances where AI generates distorted information, impacting users' understanding and trust.

Research Objectives

• Systematically categorize distorted information in artificial intelligence-generated content (AIGC).

• Conducted a case study with ChatGPT, analyzing 243 instances of distorted information.

Key Findings

• Identified 8 first-level error types:

  • Overfitting

  • Logic errors

  • Reasoning errors

  • Mathematical errors

  • Unfounded fabrication

  • Factual errors

  • Text output errors

  • Other errors

• Each type is further subdivided into 31 second-level error types.

Practical Implications

• The classification aids users in spotting distorted information.

• Assists developers in enhancing AI-generated tools' quality.

AI and Distortion

• Distortion can occur as disinformation (intentional) or misinformation (unintended).

• The phenomenon raises concerns regarding the authenticity of information in AI applications across sectors like healthcare and media.

Future Directions

• Continuous refinements to classification necessary due to evolving AI models.

• Further research needed into distortions from other AIGC models beyond ChatGPT.