AI and the Challenge of Left-Handed Writing

AI is revealing its limitations in unexpected areas, such as illustrating a person using their left hand to write. This revelation came from Prime Minister Narendra Modi during his address at the AI Action Summit held in Paris on Tuesday, amidst a landscape overwhelmed with AI-generated deepfakes.

As he co-chaired the summit alongside French President Emmanuel Macron, PM Modi initiated his discussion with an “experiment.”

He stated, “When you upload your medical report to an AI application, it can interpret it in straightforward language without any complex terms, providing insights into your health.” This demonstrates typical AI capabilities—no surprises there.

However, he then offered an intriguing insight: “If you request that same application to produce an image of a left-handed individual writing, it is most likely to depict someone writing with their right hand.” The reasoning behind this is that AI systems predominantly learn from data showcasing right-handed individuals.

This raises an important question: Is AI still encountering this issue in 2025, following significant technological advances? To investigate, we conducted a straightforward experiment using the popular AI model ChatGPT.

When prompted to generate a visual of a person writing with their left hand, the AI defaulted to producing an image of a right-handed writer. Even upon request for an illustration of an individual writing with both hands, the outcome remained focused on the right hand. This experience supports the Prime Minister’s assertion about AI’s difficulty with left-handed representations.

Thus, it is evident—AI indeed struggles with illustrating left-handed writing. But what lies behind this phenomenon?

Understanding AI Bias

The Prime Minister highlighted a notable example indicating that AI models mirror human biases. Since the majority of the global population is right-handed, AI is usually trained on extensive datasets filled with images of right-handed individuals. This is why, when instructed to create a left-handed writer, AI still resorts to the right hand.

Moreover, this issue of bias is complex and not limited to just handedness; it can lead to flawed generalisations related to race, gender, and other characteristics.

AI systems have errantly categorised individuals with darker skin tones as non-human. Likewise, when tasked to depict a doctor, AI frequently defaults to a male portrayal, reflective of the predominant biases in the training material.

The Importance of AI Ethics

This highlights why AI ethics is a significant subject of discussion. Although developers strive to mitigate bias, completely eliminating it is nearly unattainable because AI learns from human-generated content that is inherently laden with biases.

PM Modi’s discussion at the AI Action Summit transcended the issue of a minor discrepancy in AI-based image generation. It served as a vital reminder that the fairness of AI is directly proportional to the quality of the data provided. For AI to genuinely represent unbiased outcomes, careful consideration and responsibility in its development and training processes are essential.