Natural Language Processing (NLP) is a branch of AI where computers are trained to understand and draw insights from text and speech. NLP technologies came into focus in the public eye in early 2023 with the release of “ChatGPT.” NLP tools offer exciting opportunities for evaluators to automate and speed-up time-consuming tasks involving tech generation and analysis. They also pose significant challenges and risks for evaluators and wider society. Because NLP technologies are usually trained on text from the internet, they perpetuate biases (racism, sexism, ethnocentrism, and more). Tools like ChatGPT are not fine-tuned for evaluation purposes, limiting their accuracy and usefulness. Other challenges and risks relate to privacy, the pervasiveness of surveillance through data, lack of transparency, data quality and validity, extractive models and problematic supply chains. To overcome these technical challenges and ethical risks, the evaluation community will need to work collaboratively to co-develop tools and ethical standards.
