ChatGPT Exhibits Bias Toward Developed Countries Over Developing Ones, as Indicated by a Sentiment Analysis Approach
Georgios P. Georgiou
Abstract
This study analyzes how ChatGPT characterizes developed and developing countries using a sentiment analysis framework. We selected 10 countries with the highest Human Development Index (HDI) and 10 countries with the lowest. The sentiment analysis provided scores indicating the degree of positivity in the descriptions of these countries provided by ChatGPT. The results revealed that ChatGPT generally expressed positive sentiments about all countries. However, strong evidence emerged showing that countries with high HDI received more positive sentiments compared to those with low HDI. These findings highlight the bias of the model in describing developed versus developing countries. Ultimately, the study highlights the importance of adjusting large language models to ensure fairer representations of countries.