Personality Traits of GPT-2, ChatGPT, and GPT-4 in Interviews

Study Details

• Data Sample: Over 6,000 interview responses generated by GPT-2, ChatGPT, and GPT-4.
• Analysis Process: Utilised Sapia.ai’s text-to-personality models to infer HEXACO personality dimensions from the generated responses.

Key Findings

• Personality Dimensions: ChatGPT scores higher in honesty/humility, agreeableness, and conscientiousness compared to GPT-2, indicating more truthful and appropriate responses.
• Model Comparisons: GPT-4 exhibits higher extraversion, conscientiousness, and openness compared to ChatGPT.

Ethical Implications: The similar personalities projected by these models in interviews raise fairness and ethical concerns, particularly for online interviews.

Key Takeaway
Generative models like GPT-2, ChatGPT, and GPT-4 exhibit distinct personality traits, influencing their perception and implications when used by candidates in online interviews. Understanding these traits is crucial for grasping the broader impact of conversational AI, particularly on hiring outcomes. This knowledge helps mitigate potential biases and maintain the integrity of the interview process at a time where generative AI is becoming more prevalent.