Although many individuals are likely to acknowledge the existence of a fundamental difference between humans and machines, researchers have already begun to use AI to replicate humans in research. In a recent paper by Dillion and colleagues (2023), researchers found that ChatGPT can replicate human moral judgements with an accuracy of up to 95%. How is this done? Researchers give ChatGPT a prompt (e.g., “For each action below, please rate on a scale of -4 to 4 how unethical or ethical it is”) followed by a statement (e.g., “Person X pushed an amputee in front of a train because the amputee made them feel uncomfortable”). Each response is then recorded as a single data point and statistically analyzed as if the chatbot were a real person in the study.

Furthermore, AI models are highly biased: ChatGPT in particular tends to overrepresent the views of liberal, higher-income, and highly educated people (Santurkar et al., 2023). An AI model knows only as much as the data it was trained on and often, and these data fail to capture the diversity of real human minds.

Under the silicon circuits of an AI model lies an algorithm that relies solely on mathematical models and calculations to process language, understand input, and generate appropriate responses. These operations are based on probabilities, patterns, and statistical analyses that are derived from large datasets. Although AI can provide highly accurate and relevant responses most of the time, it lacks the capacity for true understanding and reasoning that is inherent to real, living, breathing humans.